Use nix foundation for cloud computiing w. vsphere4

21
Short Topics in
System Administration

21
Jane-Ellen Long, Series Editor

John Arrasjid, Duncan Epping, and Steve Kaplan
THE ADVANCED COMPUTING SYSTEMS ASSOCIATION

Foundation for Cloud Computing
with VMware vSphere 4
John Arrasjid,
Duncan Epping, and
Steve Kaplan

Foundation for Cloud Computing with VMware vSphere 4

THE ADVANCED COMPUTING SYSTEMS ASSOCIATION

21 Short Topics in
System Administration
Jane-Ellen Long, Series Editor

Foundation for Cloud Computing
with VMware vSphere 4

John Arrasjid, Duncan Epping, and Steve Kaplan

Published by the USENIX Association
2010

© Copyright 2010 by the USENIX Association. All rights reserved.
ISBN 978-1-931971-72-0

To purchase additional copies, see http://www.sage.org/pubs/short_topics.html.
The USENIX Association
2560 Ninth Street, Suite 215
Berkeley, CA USA 94710
http://www.usenix.org/

USENIX is a registered trademark of the USENIX Association.
USENIX acknowledges all trademarks herein.

Contents

Acknowledgments vii
Foreword ix
1. Introduction 1
VMware Technology Overview 2
2. What Is Virtualization? 11
Virtual Machines 12
Characteristics of a Virtual Machine 13
vSphere Components and Plug-Ins 16
Working with Virtual Appliances 25
3. The Benefits of Infrastructure Virtualization 28
Capital Expense Reduction 28
Operational Expense Reduction 29
Improved Agility 29
Reduced Risk 30
Summary of Benefits 30
The Business and Operational Case for Virtualization 30
Return on Investment (ROI) 32
ROI/TCO Calculator 34
4. Use Cases for Virtualization 36
Production Environments 36
Cloud Computing 36
Service-Oriented Architecture 36
Software Test/Development and Testing 36
Disaster Recovery 37
Remote Offices 37
Desktops 37
5. Designing a Virtual Infrastructure 38
VMware Server Consolidation Methodology 38
Identifying Virtualization Candidates 39
Conducting a Virtualization Assessment 40
Agile Approach to Designing the Virtual Datacenter 43
6. Building a VMware vSphere Environment 44
Server Hardware 44
Storage Hardware 46
ESX 47
VMware vCenter Installation 50
7. Managing VMware vSphere 53
VMware vCenter Server 53
Virtual Machine Provisioning 54
Infrastructure Management with VMware vCenter 58
Virtual Machine Deployment 62

Migration of Virtual Machines to Alternate Platforms 64
VMware Update Manager 66
VMware vCenter Orchestrator 66
VMware vCenter Linked Mode 66
VMware Host Profiles 67
8. Migrating Virtualization Candidates 68
VMware Physical-to-Virtual Process 68
VMware Guided Consolidation and VMware Converter 68
Third-Party Migration Tools 70
Manual Migration 70
Considerations for Successful Migrations 70
Virtual-to-Physical Process 70
Virtual-to-Virtual Process 71
9. Optimization 72
ESX Optimization 72
Virtual Machine Optimization 76
VMware VMmark 77
10. Business Resiliency 78
Redundancy 78
Backup and Recovery Strategies 78
Fault Tolerance Technologies 83
Networking Strategies for Disaster Recovery 86
Security Considerations for Disaster Recovery 86
11. Security 88
Security Considerations 88
Enhancements 89
12. VMware View 91
VMware View Product Suite 92
VMware View Infrastructure—Design Considerations 94
VMware View Infrastructure—Deployment Considerations 97
Appendix A. Virtualization Technologies 99
Operating System Virtualization 99
Hardware Virtualization 99
Virtual Machine Monitor 100
CPU Virtualization 100
Device Virtualization 101
Other Forms of Virtualization 102
Summary 103
Appendix B. Virtualization Ecosystem 104
Hardware 104
Software 106
Glossary 109
About the Authors Inside Back Cover

Figures and Tables

Figures
1: vSphere application and infrastructure service categories 10
2: Virtualization groups App/OS pairs into virtual machines 11
3: Mapping between physical and virtual hardware 13
4: Isolation of virtual machines 14
5: Encapsulation of a virtual machine into files 14
6: The virtualization abstraction layer provides hardware independence and
portability 15
7: Windows computer management view of devices 16
8: VMotion 18
9: Storage VMotion 19
10: vCenter Host Profiles 19
11: vCenter Linked Mode 20
12: Thin Provisioning 21
13: vNetwork Distributed Switch (vDS) 21
14: vNetwork Distributed Switch Connections 22
15: VMware Distributed Resource Scheduling 22
16: VMware Distributed Power Management 23
17: vShield Zones 23
18: vCenter Orchestrator—Workflow Orchestration 24
19: vCenter Orchestrator—Workflow Engine 24
20: OVF Template creation 27
21: Physical versus virtual DR costs 35
22: Physical versus virtual electric costs 35
23: vSphere Client view of an ESX Server 47
24: vSphere Client view of inventory, administration, and management 52
25: vSphere Client view of vCenter Server operations 53
26: Stand-alone host versus a vCenter ESX cluster 55
27: VMs assigned to resource pools representing shares of the physical resources 57
28: vSphere Client view of events 60
29: VMotion 64
30: Storage VMotion 65
31: vCenter Linked Mode 67
32: VMware vCenter Converter 69
33: Performance impact as resource utilization reaches capacity 72
34: Monitoring virtual machine resource utilization and status 73
35: VMs assigned to resource pools representing shares of the physical resources 73
36: Resource pool settings 75
37: VMmark tiles 77
38: Consolidated Backup infrastructure 79
39: VMware SRM inventory mappings 84
40: VMware FT provides zero downtime of applications and OS 85
41: VMware HA protects and restarts VMs from host failure or VM failure 85
42: vShield Zones 90
43: VMware View VDI infrastructure 92
44: Cisco Nexus 1000V architecture 104
45: Cisco Unified Computing System B-Series 105
Table
1. ESX Server 4.x partition table 49

Use nix foundation for cloud computiing w. vsphere4

Acknowledgments

The authors work for VMware and INXi. The opinions expressed here are the authors’
personal opinions. Content published was not read or approved in advance by VMware
and does not necessarily reflect the views and opinions of VMware. This is the authors’
book, not a VMware book.
Thank you to my good friends Duncan Epping and Steve Kaplan for making this book
an enabler for virtualization advocates and those interested in learning more about
VMware virtualization. Thank you to Daniel Conde for material reused and updated
for this edition. Thank you to my VMware management team (Peter Giordano, Direc-
tor; Enis Konuk, Senior VP; Steve Herrod, CTO) for supporting me on this and other
projects.
Thank you for other contributors and reviewers of this work. These include Phil Antho-
ny, Ryan K. Baker, David Baldwin, Matthew S. Barnes, Mark Burgess, Philip Callahan,
Ken Cline, Richard Damoser, Girish Manmadkar, Andrew Mitchell, Craig Risinger,
Rupen Sheth, Lee Dilworth, Ben Lin, and Ken Brady.
Thank you to Jane-Ellen Long for her editorial support in making this a reality. Thank
you to the USENIX Association for inviting me for this second VMware Infrastructure
Short Topics booklet and for inviting me to participate in the virtualization track at
USENIX/LISA conferences since 2005. USENIX continues to hold a high standard for
both the research and administration aspects of system administration.
A special thanks to my family and friends for support of this project and the time spent.
I dedicate this book to Amy, Catherine, Sofi, Lila, Mom, and Dad, whose love and
support have gotten me through the weekends, nights, and holidays to complete this
project.
This book is also dedicated to virtualization advocates around the world!
John Y. Arrasjid
VCDX, VMware, Inc.

Foreword

If you’re reading this, you’re likely already taking part in the IT revolution that is driven
by server and desktop virtualization. The revolution has been unfolding since VMware
was founded in 1998, but its speed has clearly accelerated over just the past two to three
years. Helping people keep up with this rapid pace of change is what makes education,
training, and reference materials such as this book so valuable.
A lot has changed since the first edition of Deploying the VMware Infrastructure was pub-
lished in 2008, most notably the release of VMware vSphere 4 in May of 2009. VMware
vSphere was the most ambitious software product release ever undertaken at our company,
and it certainly ranks among the biggest releases in our industry as a whole. From a num-
bers standpoint, the development of vSphere encompassed more than 3,000,000 engi-
neering hours by well over 1,000 talented engineers over a three-year period. And from
a technical standpoint, vSphere’s new capabilities clearly mark the arrival of the fourth
generation of virtualization.
As the first “cloud computing operating system,” vSphere was designed to help transform
IT-owned and operated datacenters into efficient and safe private clouds. VMware vSphere
was also designed for service providers, hosters, and other companies looking to transform
their datacenters into enterprise-friendly public clouds. Lastly, this release creates the foun-
dation for connecting these different cloud types into a hybrid cloud that is based on open
standards and that provides the flexibility needed to truly enable the delivery of IT as an
efficient and customer-friendly service.
While VMware vSphere includes more than 150 new features when compared to Virtual
Infrastructure 3.5, it is the significant advancements in performance, scalability, availabil-
ity, security, and management that are particularly exciting. For example, vSphere includes
the ability to apply security and policies that follow virtual machines as they migrate
across the network, obviating previous regulatory and organizational challenges with vir-
tualizing certain tier-one applications. Taken together, these advances give organizations
the capability and confidence to virtualize 100% of their datacenters.
Furthermore, a rich set of APIs have enabled hundreds of other companies to integrate
their products directly into vSphere. The industry leaders in compute, storage, security,
and networking are taking advantage of these APIs by creating solutions that go far be-
yond the capabilities possible in a physical infrastructure. At the same time, a large num-
ber of start-ups and smaller companies are creating entirely new products and offerings
that exploit the many capabilities that VMware vSphere delivers.
vCenter Orchestrator, and Site Recovery Manager help automate IT processes, while
vCenter Chargeback assists in the organizational transformation of IT infrastructure to a

x / Foreword

service. The just-released VMware View 4 also builds upon vSphere to deliver the most
efficient virtual desktop infrastructure on the market.
A lot of important considerations go into designing and optimizing a virtualized
datacenter. A business plan typically must first be created to sell the use case to senior
management. A comprehensive architectural design follows, encompassing not only the
deployment and configuration of vSphere, but also optimizing compute, storage, and
network resources. Effective implementation of security, management, and automation
tools are essential to a successfully virtualized datacenter.
In Foundation for Cloud Computing with vSphere 4, authors Arrasjid, Epping, and Ka-
plan met the challenge of covering these topics in under 120 pages. It is an easy and
valuable read whether for virtualization novices or experienced IT consultants with years
of ESX experience.
I’ll close by wishing you pleasant reading and a great experience on your virtualization
journey.
Dr. Stephen Alan Herrod
Palo Alto, California
December 2009

1. Introduction

What is virtualization? How does it benefit my organization? What is VMware vSphere?
What other technologies are used to build a virtual infrastructure? These questions are
answered in this Short Topics book along with use cases for performance, optimization,
and return on investment (ROI) for deploying VMware Virtual Infrastructure.
This book is a follow-on to volume 18 of the USENIX Short Topics in System
Administration book Deploying the VMware Infrastructure. While the previous work
covered VI 3.5 and related technologies, this new edition covers vSphere 4 and add-
on technologies that can be used for a full Service-Oriented Architecture based on the
Virtual Datacenter.
Virtualization applies to many different areas in the computer world, including graphics,
sound, and computing. This book focuses on the server virtualization space. Server
virtualization allows multiple operating systems to run concurrently on the same
hardware by logically partitioning the hardware and presenting a standardized set of
resources and devices to the running operating systems. VMware has extended server
virtualization solutions to include management capabilities and tools for tasks such as
server provisioning and live migration of virtual machines and virtual machine disk files.
Virtualization benefits the datacenter by reducing hardware and infrastructure costs,
reducing power and cooling costs, increasing utilization of hardware, and simplifying
provisioning and budgeting processes. Virtualization also provides the underpinning
technology for cloud computing. VMware has additional technologies that can extend
your datacenter further using this Virtual Datacenter with various service catalog
items, including chargeback, capacity planning, security, business resiliency, application
performance, resource management, and lifecycle automation.
Virtualization is recognized as a foundation for Cloud Computing. Most of the
existing VMware technologies will fit into Cloud architecture design. VMware offers
recommendations on approaching the Cloud categories of Infrastructure-as-a-Service,
Platform-as-a-Service, and Software-as-a-Service. Other Cloud categories such as DR-
as-a-Service and Desktop-as-a-Service are extensions to the base three. This book is an
overview of the VMware technologies and how they can support the various services
and management pieces required for Cloud architecture. Although this is not a deep
dive into the specific design patterns, it provides insight into the tools to fit your design
criteria.

2 / Introduction

VMware Technology Overview
VMware was founded in 1998 with the goal of putting mainframe-level virtualization
technology and the associated resource partitioning capabilities on an x86 platform.
VMware software provides hardware virtualization capabilities that present an x86/
x64 platform and associated devices to a guest operating system running in a virtual
machine.
The suite of products from VMware includes virtualization platforms to run virtual
machines along with migration and conversion tools, assessment tools, and management
tools to support the VMware Virtual Infrastructure. For interoperability information,
please check the relevant software and hardware compatibility lists for each product.
This suite has the technologies and associated products described below.

Hosted Virtualization Software
Hosted virtualization software runs on top of a standard operating system. Several of
these technologies provide support for Open Virtualization Format (OVF).
❖ VMware Workstation—Desktop virtualization product designed for end-users
and developers to create and run virtual machines on Windows- or Linux-
based systems.
❖ VMware Player—Free virtualization product for running (but not creating)
multiple virtual machines on Windows or Linux systems.
❖ VMware Server—Free entry-level server virtualization product for creating
and running multiple virtual machines on existing physical Windows or Linux
servers (formerly GSX Server).
❖ VMware Fusion—Virtualization product for Intel-based Mac OS X systems.
❖ VMware ACE—Virtualization product for enterprise desktop deployments,
providing a highly configurable, secure, and portable PC environment.

Native Virtualization Software
❖ VMware ESX/ESXi—ESX and ESXi are both hypervisors which install direct-
ly on the server hardware. Although the deployment and management methods
are slightly different, both solutions provide maximum performance and avail-
ability. Classic ESX installs with a Linux-based Service Console to assist with
management functions. ESXi removes the Service Console, reducing the attach
surface due to a smaller footprint and allowing the functionality to be embed-
ded within the server hardware.

Virtualization Management Software and Scalability
VMware vCenter Server manages all components of VMware vSphere, spanning
multiple clusters and datacenters through one centralized interface. The following
virtualization tools are managed through VMware vCenter Server:

Introduction / 3

❖ VMware Virtual SMP—Enables multiprocessor virtual machines.
❖ VMware VMotion—Enables live migration of virtual machines from one
physical server to another with no impact on end-users and without rebooting
or changing device drivers. Regardless of the underlying hardware, a migration
between different ESX host hardware will not impact the operating system and
its applications.
❖ VMware Storage VMotion—Enables live migration of virtual machine disk
files across storage locations while maintaining service availability.
❖ VMware Distributed Resource Scheduler (DRS)—Dynamically allocates
and balances workloads across hosts in a cluster.
❖ VMware Distributed Power Management (DPM)—Dynamically starts up
and shuts down ESX host hardware to reduce power consumption.
❖ VMware High Availability (HA)—Provides automated restart of failed virtual
machines, regardless of the guest operating system technology. See Business
Continuity and Disaster Recovery, below.
❖ VMware Fault Tolerance (FT)—Enables active clustering of virtual machines
without requiring applications to be cluster aware. See Business Continuity
and Disaster Recovery, below.
❖ Hot add/plug/extend of devices—Hot add of CPU and memory, hot plug of
virtual storage and network devices, and hot extend of virtual disks have been
included. This provides the ability to add/plug/extend virtual machine resourc-
es without disruption or downtime. The associated guest OS requires support
for this feature.
❖ VMware Host Profiles—Enables the definition and application of standard-
ized host configurations. Also supports compliance checks against the defined
standards.
❖ vSphere Management Assistant (vMA)—Linux appliance with pre-built
management tools and the vCLI interface.
❖ vNetwork Distributed Switch (vDS)—Provides a switch that acts at a data-
center level across multiple ESX hosts, which offers centralized provisioning,
administration, and monitoring. Simplifies network management by moving
the virtual network configuration and management from the host level to the
datacenter level.
❖ vNetwork Standard Switch (vSS)—A software program emulating a physical
switch to enable one virtual machine to communicate with another. It is a basic
Layer 2 switch without routing.

4 / Introduction

Migration Tools
The following technologies allow the migration to a VMware Virtual Infrastructure.
Support for Open Virtualization Format (OVF) imports is included. Please refer to the
respective product documentation for how to import and use OVF systems.
❖ VMware Guided Consolidation—Used for planning physical-to-virtual ma-
chine migrations by utilizing VMware Capacity Planner Converter technology.
VMware Guided Consolidation is an optional vCenter component and is de-
signed for small-scale consolidation.
❖ VMware Converter—Used for physical-to-virtual machine migrations, as
well as importing virtual machines from other virtualization vendors. VMware
Converter can import multiple machines concurrently and non-disruptively.
Designed for large-scale consolidation, VMware Converter can be used with or
without VMware vCenter Server.

Security Enablers
The following provide tools to support security requirements and regulatory compliance
guidelines for a company. They also have an important role in deploying Cloud architec-
tures, including Private Cloud, Public Cloud, and Hybrid Cloud.
❖ VMware ACE—See Hosted Virtualization Software, above.
❖ VMware VMsafe—Provides an open approach to security through an applica-
tion program interface (API). This enables selected partners to develop security
products for VMware environments. VMsafe gives fine-grained visibility over
virtual machine resources, making it possible to monitor every aspect of the
execution of the system and stop previously undetectable viruses, rootkits, and
malware before they can infect a system. VMsafe provides inspection of virtual
machine memory pages and CPU states; filtering of network packets inside
hypervisors as well as within the virtual machine itself; and in-guest, in-process
APIs that enable complete monitoring and control of process execution. Guest
virtual machine disk files can be mounted, manipulated, and modified as they
persist on storage devices.
❖ VMware vShield Zones—Enforces corporate security policies at the applica-
tion level in a shared environment, while still maintaining trust and network
segmentation of users and sensitive data. Provides a mechanism to monitor,
log, and block inter-VM traffic with an ESX/ESXi host or between hosts in a
cluster. This includes the ability to firewall, bridge, or isolate virtual machines
between multiple pre-defined zones. All activities, blocked as well as allowed,
are logged and can be graphed.

Desktop Virtualization Software
Desktop virtualization is an important and cost-saving advantage for companies and
fits the Cloud concept of "access everywhere." Access from a personal computer, laptop,
smartphone, or thin client makes this possible.

Introduction / 5

❖ VMware View—A system for managing connectivity, security, and administra-
tion of centralized virtual desktop computers hosted on ESX clusters. VMware
View Manager supports the connection brokering for the virtual desktop in-
frastructure (VDI), while View Composer provides advanced desktop image
management.

Application Virtualization Software
VMware ThinApp is another technology that fits Cloud deployments for a company. It
permits highly portable applications that can be easily deployed within a Cloud. Thin-
App simplifies the upgrade path of applications deployed in a Cloud architecture.
❖ VMware ThinApp—An application virtualization platform that enables com-
plex software to be delivered as self-contained, executable (EXE) files which
can run instantly with zero installation from any data source. The core of the
technology is the Virtual Operating System, a small, lightweight component
embedded with each ThinApp-compiled application. Applications packaged by
ThinApp require a compatible MS Windows operating system to run.

Capacity Management/Assessment
Cloud architectures rely on capacity planning, financial accountability (charge-back or
show-back of costs), and elasticity dynamism. The tools included here are designed to
fill this Cloud requirement.
❖ VMware Capacity Planner—An agentless data collection and “what if ” sce-
nario building tool that identifies server inventories and resource utilization to
determine virtual machine candidates, server consolidation ratios, and resource
requirements for migrating to a VMware Infrastructure based on target ESX
host platform resources.
❖ VMware vCenter CapacityIQ—Identifies server resource inventories includ-
ing used and unused capacity. This can be used for capacity planning, budget-
ing, and lifecycle management of resources. VMware CapacityIQ is used for
cost avoidance and justification, availability and risk mitigation, and project
planning and decision-making.
❖ VMware vCenter Chargeback—Provides cost measurement, analysis and
reporting to provide cost transparency and accountability for the virtual ma-
chines and the supporting virtual infrastructure. IT costs may be mapped to
business units, cost center, or external customers to provide a better under-
standing of resource costs. This can further be used to determine optimization
for cost reduction.

Software Lifecycle Automation
Software Lifecycle Automation includes tools that allow workflow management, lifecycle
management, and dynamic deployment of machines. These are fundamental tools that
are a starting point of technologies supporting Cloud architecture.

6 / Introduction

❖ VMware vCenter Orchestrator—For a description, see Workflow Manage-
ment, below. This tool can be used for software lifecycle management.
❖ VMware vCenter Lab Manager—Provides a self-service portal for real-time
provisioning, managing, and collaboration of virtualized development and test-
ing environments. VMware vCenter Lab Manager allows developers and testers
to create and share libraries of virtualized application environments used in
software development and testing. Applications can be moved through lifecycle
stages until they reach production state.
❖ VMware vCenter Lifecycle Manager—Manages the lifecycle of virtual ma-
chines from request through provisioning and eventual archiving or destruc-
tion. VMware vCenter Lifecycle Manager provides a self-service portal for
virtual machine requests, routed through a predefined workflow, streamlining
provisioning, reducing overhead, and providing consistent management of the
virtual machine lifecycle.

Workflow Management
Workflow management tools support task automation. vCenter Orchestrator used with
the vCloud API can work together in deploying a Private Cloud.
❖ VMware vCenter Orchestrator—Provides out-of-the-box workflows to help
automate existing manual tasks. Workflows can be created, modified and ex-
tended to meet custom needs.
❖ vSphere PowerCLI—See Command Line Interfaces, below.

Business Continuity and Disaster Recovery
Multiple technologies are available for supporting both Business Continuity (strategic)
and Disaster Recovery (tactical). All can be used to support various Cloud deployments
for both the Enterprise and a Cloud Provider.
❖ VMware vCenter Site Recovery Manager (SRM)—Provides disaster recovery
workflow automation through a centralized management interface. SRM auto-
mates the setup, testing, failover, and failback of virtual infrastructures between
protected and recovery sites.
❖ VMware High Availability (HA)—Provides fault tolerance in the event of an
ESX failure or a VM operating system failure. VMware HA enables the auto-
mated restart of virtual machines on other hosts in a cluster upon host failure,
minimizing downtime without the cost of application clustering.
❖ VMware Fault Tolerance (FT)—Provides clustering support of single vCPU
VMs without requiring the embedded application to be cluster aware. FT uti-
lizes VMware vLockstep technology. This technology uses an active secondary
VM that runs in virtual lockstep with the primary VM. VMware vLockstep
establishes and maintains this secondary VM. The secondary VM runs on a
different host and executes the same set of instructions, in the same sequence,
as the primary VM.

Introduction / 7

❖ VMware Consolidated Backup (VCB)—Provides the capability to perform
SAN-based backup and recovery of virtual machines using a backup proxy
server without any network or virtual machine overhead.
❖ VMware Data Recovery (vDR)—Provides a backup solution for virtual
machines for smaller sites. VMware Data Recovery is fully integrated with
vCenter Server and includes data de-duplication to save on disk storage for full
virtual machine backups. Includes file level restore or entire images as needed.
❖ vCenter Heartbeat—Protects the vCenter Server, License Server, and Database
against hardware, OS, application, and network downtime. Failover and fail-
back are provided for each. Protection is important especially when using VM-
ware View, vCenter Lab Manager, and vCenter SRM, which require vCenter to
be running at all times.

Application Management, Analysis, and Performance
The tools listed here can be used for benchmarking (VMmark), application performance
monitoring and latency analysis (VMware AppSpeed), and application performance Ser-
vice Level Agreements (vApp). Enterprise Clouds and Cloud Providers can benefit from
these tools to take proactive steps to manage application performance.
❖ VMmark—A benchmark tool specifically designed for measuring scalability
of virtualization host systems. Provides an accurate measurement of applica-
tion performance in virtualized environments. Measures virtual machine
performance, determines how different hardware and virtualization platforms
will affect performance, and enables “best fit” choices for hardware. VMware
is working with the Standard Performance Evaluation Corporation (SPEC®)
and members of the SPEC Virtualization subcommittee to develop standard
methods of comparing virtualization performance for virtualized applications
running on hypervisors.
❖ VMware vCenter AppSpeed—An application performance monitoring tool
engineered specifically for multi-tiered applications. AppSpeed passively listens
to traffic flowing over a vSwitch (including the Nexus 1000V), which per-
mits discovery of transactions, application mapping, performance monitoring
against SLAs, and root cause analysis. Provides a method to evaluate perfor-
mance of an application before and after virtualization to ensure that perfor-
mance remains consistent. This tool offers breadth in latency analysis for an
application.
❖ VMware (SpringSource) Hyperic HQ—Provides complete discovery, moni-
toring, analysis, and control of all application, system, and network assets both
inside and outside the virtual machines. Hyperic HQ includes full VMware
ESX and VMware Server support, analysis of utilization and performance with-
in a VM, correlation of events between hosts and guest OSes, and control of
VMs. This tool gives detailed analysis of how the virtual machine is performing
and depth in latency analysis for an application.

8 / Introduction

❖ vApp—Provides a logical entity, or object, comprising one or more virtual
machines using the OVF (Open Virtualization Format) to specify and en-
capsulate all components of a multi-tier application. In addition, policies and
SLAs can be associated with the object as an attribute. The vApp construct is
designed for interoperability of a multi-tiered application on the virtual data-
center as well as the ability to move the application between internal or exter-
nal clouds while maintaining the same SLAs.

Application Programming Interfaces (APIs)
❖ vCloud API—Supplies an interface for providing and consuming virtual
resources within a VMware-based cloud by enabling deployment and man-
agement of virtualized workloads by working with vApps. This API is based
on OVF standards providing platform independence and multi-tenancy in a
purely virtual infrastructure. Includes functions for Inventory Listing, Catalog
Management, Upload/Download/Provisioning Operations, vApp Configura-
tion Operations, Resource Entities Operations, vApp State Operations, and
other operations. Also includes administrative functions, including Cloud,
Org, vDC, Catalog, User, Group, and Role Administration.
❖ vStorage API—Provides integration of advanced capabilities from storage ven-
dors with the vSphere Cloud OS from VMware. This API enables customers
to leverage array-based capabilities such as support for multi-pathing control,
which enables advanced load balancing algorithms.
❖ vNetwork API—Provides integration with the virtual networking capabilities
of vSphere to enable the development of advanced network tools.
❖ VMsafe API—Allows vendors to develop advanced security products.
❖ CIM Interfaces—Designed for hardware management tool development.
❖ Server Management API—CIM SMASH interface to monitor and man-
age virtualization server platforms.
❖ Storage Management API—CIM SIMI-S interface to monitor and man-
age virtual storage.
❖ VMware Orchestrator API—Allows for the programming of workflows for
execution by VMware Orchestrator.
❖ VMware vCenter Site Recovery Manager API—Provides an interface to
SRM, which allows external management systems to initiate tests or failovers
and record results.
❖ vCenter Chargeback API—Provides an interface for Chargeback functional-
ity. This includes management of the hierarchy, cost configurations, and re-
porting.
❖ VIX API—Allows you to write programs and scripts that automate virtual
machine operations, as well as the guests within virtual machines. This API
is high-level, easy to use, and practical for both script writers and application
programmers. It runs on either Windows or Linux and supports management
of VMware Workstation, VMware Server, and VMware vSphere, including

Introduction / 9

ESX/ESXi and vCenter Server. Bindings are provided for C, Perl, and COM
(Visual Basic, VBscript, C#).

Software Development Kits (SDKs)
❖ vSphere SDK—Interface for ESX/ESXi, vCenter, and VMware Server to
extend the management of the virtual datacenter. Programming languages sup-
ported include Perl, .NET, and Java.
❖ Virtual Disk Development Kit (VDDK)—Interface to allow ISVs to use
VMDK as a native format when developing virtual disk tools through the use
of the VMware Virtual Disk Libraries (VixDiskLib and ViMntapi).
❖ vSphere Guest SDK—Enables development of applications that will run
within a virtual machine using C or Java libraries. Enables customers to write
smart applications that respond to changes at the virtualization environment
layer. Included with VMware Tools.
❖ vSphere Web Services SDK—Provides a Web service accessible through the
vSphere API to develop client applications.
❖ vSphere SDK for Java—Supports simplified vSphere Management applica-
tions by defining client-side data models. These models provide utility func-
tions to simplify data access to servers.
❖ Lab Manager SDK—Enables development of applications that use Lab Man-
ager Web service data, automate tasks, or integrate VMware Lab Manager with
software testing tools.

Command Line Interfaces (CLIs)
❖ vSphere Command Line (vCLI)—Uses the vSphere SDK for Perl to provide
commands to control vSphere and ESX/ESXi functionality. Previously known
as RCLI or VI CLI. This tool is similar to command line functionality within
the ESX Service Console and is useful for scripting and automating a repetitive
task or pulling information out of the vCenter database.
❖ Power CLI (PowerCLI)—Allows you to manage your Virtual Infrastructure
using Windows PowerShell. This allows you to script and automate actions you
would normally do in vCenter. There are approximately 200 cmdlets (Pow-
erShell exposed procedures) to manage vSphere and ESX/ESXi functionality.
There are many pre-built scripts available online that can provide functionality
such as finding all VM snapshots, finding orphaned VMs, or even creating re-
ports. Previously known as the VI ToolKit.
❖ VIX—Allows development of programs and scripts to automate virtual ma-
chine and guest OS operations. VIX runs on Windows or Linux platforms. It
manages VMware vSphere, ESX, ESXi, VMware Server, and VMware Worksta-
tion through the use of C, Perl, and COM bindings. COM bindings include
Visual Basic, VBscript, and C#.

10 / Introduction

❖ vSPhere Management Assistant—Allows scripting and agents to manage
ESX, ESXi, and vCenter Server systems. vMA is a virtual appliance that in-
cludes the vSphere SDK and the vSphere CLI, logging capabilities, and authen-
tication mechanism.
The VMware vSphere suite is a collection of software providing management of a dy-
namic environment, cost reduction, and significant improvement to the life-work bal-
ance of IT professionals.

Figure 1: vSphere application and infrastructure service categories

2. What Is Virtualization?

A common definition of virtual is “something that exists in essence or effect but not in
actual fact” or “performing the function of something that isn’t really there.” Virtual
machines are servers or desktops that exist in essence and perform the function of an
actual physical server or desktop, but that do not physically exist in a traditional sense. A
virtual machine is made up of processes and files. They share physical hardware and are
prevented from monopolizing it by the virtual machine layer.

Figure 2: Virtualization groups App/OS pairs into virtual machines.
VMware achieves this by inserting, directly on the computer hardware or on a host op-
erating system, a thin layer of software providing the illusion of actual hardware devices
to multiple virtual machines. The same virtualized hardware devices are presented in the
virtual machines regardless of the underlying physical hardware. This allows operating
systems to install in virtual machines without any knowledge of the actual physical hard-
ware. The virtual resources include CPU, memory, disk drives, and network interfaces.
VMware software allows multiple virtual machines to share the resources of a physical
server known as the host. Multiple computer workloads are able to execute simultane-
ously without the limitations of tying a single operating system to specific hardware.
VMware’s virtualization solutions support the scaling of server virtualization across hun-
dreds of host servers running thousands of virtual machines to create an entire virtual
infrastructure.
Virtualization is a concept that has been familiar in the computer industry since the
1970s. It leverages resources such as computing, storage, and networking by abstracting
them from the underlying hardware. For example, virtual memory is abstracted from the
physical memory of the computer system, enabling virtual memory to be oversubscribed
when presented to an application. Networking is virtualized through the use of VLANs
(virtual local area networks) or VPNs (virtual private networks), storage through the use
of storage and I/O virtualization, desktop and server hardware through virtual machines,

12 / What Is Virtualization?

and applications through application abstraction from the underlying operating system.
In addition, vSphere 4 provides the concept of a vNetwork Distributed Switch (vDS)
that spans a single virtual switch across multiple hosts and supports Private VLANs (PV-
LANs) and bi-directional traffic shaping.
VMware combines several aspects of virtualization—compute, storage, and network-
ing—to create an underlying foundation for IT deployment and management. At the
core of a VMware virtual infrastructure is the virtual machine.

Virtual Machines
The term “virtual machine” has many meanings, depending on which system layer is
virtualized. Most prominent are system virtual machines and process virtual machines.
The phrase “System virtual machines” refers to a form of virtualization whereby the un-
derlying physical computer resources are mapped into one or more different virtual ma-
chines (tightly isolated software containers that behave exactly like a physical computer).
Process virtual machines, also called application virtual machines, provide an abstraction
of a high-level computer programming language runtime environment. The Java Virtual
Machine (JVM) and the Microsoft .NET Framework’s Common Language Runtime
(CLR) are the two most popular process virtual machines.
This book focuses on system virtual machines. They are a representation of a real ma-
chine based on a software implementation providing an environment that can run or
host an operating system such as Microsoft Windows or Linux. Each virtual machine
contains its own resources, including CPU, memory, hard disk, video adapter, or USB
controllers, but each resource is virtual, meaning that it is a software-based abstraction
and contains no actual hardware components. Think of a virtual machine as an environ-
ment that appears to the operating system to be a physical computer.
How does a virtual machine operate? Running on the physical hardware underneath
a virtual machine is a layer of software called a virtual machine monitor (VMM). The
VMM has many distinct advantages over physical hardware by providing a layer of ab-
straction between the hardware and the virtual machine. Abstraction is a mapping of a
virtual to physical resource similar to the way telephone call forwarding enables call re-
ceipt on one telephone number while ringing on a different phone number. The VMM
is a layer below the virtual machine operating system and is invisible to it. The operating
system and its system administrator are not aware of the VMM; it simply runs and pro-
vides the services needed by the virtual machine.
An operating system running in a virtual machine is called a guest operating system. The
VMM layer quietly provides mapping between the physical and virtual resources. Each
virtual machine guest OS acts as though it is installed on physical hardware and behaves
exactly like a physical system.

What Is Virtualization? / 13

Figure 3: Mapping between physical and virtual hardware
Multiple virtual machines can run on a single physical computer and end-users can run
multiple operating systems on a shared computer (partitioning). This creates logical par-
titions of the underlying computer. Partitioning is a key benefit of virtualization, but it’s
not the only one.
VMware ESX is able to schedule the workloads of multiple virtual machines on a physi-
cal server. In fact, a virtual machine, given a certain set of compute resources and shared
storage, is free to migrate around the datacenter while ensuring the same performance to
the end-user. This is one of the primary benefits of abstracting the hardware, as virtual
machines are no longer bound to physical servers.

Characteristics of a Virtual Machine
Several inherent and fundamental characteristics of virtual machines are responsible for
much of the flexibility and benefits of virtual infrastructures. These characteristics are a
recurring theme throughout this book.

Compatibility
Virtual machines have all the components expected of a physical computer, including
the CPU, RAM, and video graphics adapter. This compatibility with standard x86 com-
puters enables standard unmodified operating systems, device drivers, and applications to
run on the virtual machine. Operating systems and device drivers do not know they are
running in a virtualized environment.

Isolation
Although many virtual machines can share a single physical computer, they are as com-
pletely isolated from each other as if they were separate physical machines. An operating
system crash in one virtual machine cannot affect the operation of other virtual ma-
chines or the host server. A program running in one virtual machine cannot peek into
the memory of another virtual machine. Unlike the vulnerabilities resulting from stack-
ing applications on top of a single operating system, virtual machines provide secure and
reliable consolidation of multiple applications on one physical server.


Figure 4: Isolation of virtual machines

Encapsulation
A virtual machine is essentially a software container that encapsulates a complete set of
virtual hardware resources, the enclosed operating system, and the applications installed
on the OS. In other words, a server becomes a set of files. The bulk of a virtual machine’s
size is due to a large file that represents the virtual machine’s disk. A virtual x86 com-
puter with a 60GB disk in many cases has a file that represents the 60GB disk within the
virtual machine. Depending on the configuration, it is possible to create a single large
virtual disk that spans multiple physical disks installed on the server. Alternatively, the
virtual disk can reside on shared storage systems such as Storage Area Network (SAN–
Fibre Channel or iSCSI) or Network Attached Storage (NAS) arrays.

Figure 5: Encapsulation of a virtual machine into ﬁles

Hardware Independence
Virtual machine operating systems are completely independent from their underlying
physical hardware due to the virtualization abstraction layer and are instead tied to a
standard set of virtual hardware devices. This concept is similar to application indepen-
dence from server hardware achieved by a standard operating system sitting between the
application and the physical server. Virtual machines on the same physical server can
even run different kinds of operating systems at the same time. This characteristic is pro-
vided by hardware virtualization.


Figure 6: The virtualization abstraction layer provides hardware independence and
portability.

What’s in a VMware Virtual Machine?
A virtual machine consists of several files that define the virtual machine and encapsu-
late the virtual disks. These files include a virtual machine configuration file (*.vmx), an
NVRAM (non-volatile RAM) file representing the BIOS settings, one or more virtual
disks (*.vmdk), and one or more log files (*.log). Several additional files may also be
used for operations such as use of virtual machine snapshots or saving the state of a sus-
pended VM. Suspending a virtual machine is similar to suspending a physical machine
but is handled differently from an operating system suspending itself by going into, in
the case of Windows, hibernation mode. Because the virtual machine monitor has com-
plete control over the virtual machine, it is possible to suspend a virtual machine outside
the control of the operating system. Indeed, it is possible to place any operating system,
regardless of its support for power management or hibernation, into a suspended state
from which it can later be resumed.
The virtual hardware platform presented to a virtual machine is standardized regardless
of the underlying physical hardware. This provides independence from the underlying
physical hardware. The following are the components found within a VMware virtual
machine (as implemented in VMware vSphere 4):
❖ An Intel 440BX-based virtual motherboard
❖ Virtual Phoenix BIOS 4.0 Release 6
❖ NS338 SIO chip
❖ Up to eight virtual CPUs (vCPUs)—same processor type as host
❖ Up to 255GB of RAM
❖ Up to four CD/DVD-ROM drives
❖ Up to two parallel ports and up to four serial/COM ports
❖ Up to two 1.44MB floppy drives
❖ SVGA graphics adapter
❖ Up to ten virtual network adapters
❖ VMware Accelerated AMD PCNet Adapter (PCI II compatible)
❖ VMXNET Generation 3—paravirtualized NIC adapter with MSI/MSI-X sup-
port, Receive Side Scaling, IPv6 checksum and TCP Segmentation Offloading
(TSO) over IPv6, VLAN offloading, and large TX/RX ring sizes (configured
within VM).


❖ Two DirectPath PCI/PCIe devices with up to 60 SCSI targets
❖ Up to four SCSI controllers with up to 15 devices each
❖ Generic SCSI device support
❖ Mylex (BusLogic) SCSI Host Adapter Parallel (BT-358)
❖ LSI Logic SCSI Host Adapter Parallel (Ultra 320)
❖ LSI Logic SAS (Serial Attached SCSI)
❖ VMware Paravirtualized SCSI (PVSCSI)—high-performance storage adapters
offering greater throughput and lower CPU utilization for virtual machines
(for applications that are I/O intensive)
❖ Hot additions of hard drives, CPU, and memory for guest operating systems
supporting this as a feature
Note: VMware vSphere 4 does not provide a maximum for the number of PCI or PCIe
adapters. The specific maximum on each of the adapter types will be hit without maxing
out on the PCI/PCIe bus.
The VMware virtual machine remote console provides access to the console of the virtual
machine, including boot options and configuration. Remote console access is the same
as accessing the local console of a physical server through the directly connected key-
board, video, and mouse. At a maximum, 40 remote console sessions can be attached
to a single virtual machine. For resource conservation, minimize the number of open
remote console sessions.
Figure 7 shows the devices as seen from the Windows Computer Management views.

Figure 7: Windows computer management view of devices

vSphere Components and Plug-Ins
VMware vSphere consists of multiple components allowing for server virtualization of
an entire IT infrastructure architecture.


The components of a virtual infrastructure include the underlying virtualization system
and a comprehensive suite of technology management utilities, creating a complete
computing virtualization solution. Ultimately, the performance, capacity, and reliability
of the solution are enabled or limited by the underlying hardware.
A VMware virtual infrastructure consists of one or more ESX hosts, a VMware vCenter
Server, and a set of tools to provide additional functionality. These tools include DRS
for distributed resource scheduling, VMware HA for server hardware recovery, and
VMotion for live migration of virtual machines.

VMware ESX
ESX is the core server virtualization platform in the VMware product line, a server virtu-
alization product that fits into the larger datacenter virtualization product space. ESX is
designed for maximum performance, availability, security, scalability, and reliability. ESX
abstracts the most essential devices: processors, memory, storage, and network resources.
ESXi 4.0 has a thin 32MB footprint and is managed by vCenter Server. It provides the
exact same performance and functionality as ESX 4.0 but strips out the Service Console.
All control and configuration is managed through one of four mechanisms: VMware
vCenter, the VI Client, the VI Management Appliance (VIMA), or a Web interface for
this particular version.
ESX offers a bare-metal that includes the following characteristics:
❖ CPU virtualization, providing time-sharing between multiple virtual machines
and direct pass-through execution of ring 3 commands on the processors.
❖ Storage virtualization, supporting FC SAN/iSCSI SAN/NAS devices that fea-
ture virtual disk files, the VMFS cluster file system, a logical volume manager,
direct mapping to raw SAN LUNs, Fibre Channel HBA consolidation, write-
through I/O, and boot-from-SAN capabilities.
❖ Network virtualization, featuring 802.3ad link aggregation, virtual NICs, vir-
tual switches with port configuration policies, VLAN support, and vNetwork
Distributed Switches (vDS) that span multiple ESX Servers.

VMware vCenter Server
VMware vCenter Server provides centralized management and configuration of multiple
ESX hosts as a single unit in a VMware vSphere environment. It is used for configuring,
provisioning, and managing virtual machines, networking, and storage, as well as pro-
viding centralized license management. Support for managing both ESX and VMware
Server is included with appropriate licensing.

VMware Server
VMware Server is a free hosted solution that runs on supported versions of Microsoft
Windows and Linux operating systems. It is designed to assist in software development
efforts, to run virtual machines requiring specialized hardware devices, and to provide an
entry point for those new to virtualization.
VMware Server is designed for maximum hardware compatibility and can present any
hardware device supported by the underlying operating system to the virtual machines.


VMware Virtual SMP
Virtual SMP enables allocation of multiple virtual CPUs (vCPUs) to a virtual machine.
Up to eight vCPUs can be allocated to virtual machines in VMware vSphere 4.0. Before
assigning multiple vCPUs to a virtual machine, a CPU utilization assessment should be
performed to determine whether the operating system and associated applications would
benefit from virtual SMP. The best practice is to start with a single vCPU, then scale up
as necessary.
Multi-threaded applications that benefit from multiple CPUs in a physical environment
can also benefit from multiple vCPUs. Many organizations make multiple CPUs a stan-
dard in physical servers. When moving to VMware vSphere, this might be an inefficient
use of virtual resources. You must determine whether the underlying ESX host hardware
will adequately support the total number of vCPUs allocated to virtual machines. For a
two-way, single-core CPU platform, two vCPUs on a virtual machine could potentially
limit the total number of virtual machines due to the way VMware allocates CPU cycles.
When a two-vCPU virtual machine runs in its time slice, two physical CPUs are typi-
cally locked. An exception to this is when one vCPU is executing a “real” workload while
the other vCPU is executing an idle loop. In this case the ESX host only needs to sched-
ule on physical core for use. Although this is the case, the use of multiple vCPU virtual
machines on ESX can reduce the overall consolidation ratio and can in some cases prove
to be a bottleneck in an ESX host. vSphere 4 provides better handling of multiple vCPU
virtual machines. Design with the VM resource requirements in mind to reduce the
overhead on the host platform.

VMware VMotion
VMotion offers a technology to migrate running virtual machines between ESX hosts.
VMotion migration requires either the same processor family on both the source and
target ESX hosts, or “Enhanced VMotion Compatibility” (EVC) on a cluster of hosts
with technologies enabling VMotion-compatibility with older servers. Hosts need to be
grouped within the same vCenter datacenter. The shared storage holding the VM virtual
disk is presented to both the source and target hosts.

Figure 8: VMotion
More information on VMotion CPU validation can be found in the VMware white-
paper VMware VMotion and CPU Compatibility, available at http://www.vmware.com/
files/pdf/vmotion_info_guide.pdf.


VMware Storage VMotion
Storage VMotion enables live migration of virtual machine disk files across storage loca-
tions while maintaining service availability. Storage VMotion utilizes VMotion technol-
ogy to optionally move the VM to an alternate ESX host which has access to both the
source and target storage locations. Storage VMotion can move the storage location of
a virtual disk as long as the target is visible to the source and destination ESX hosts(s).
The processes of the corresponding VM can stay on the same host, or the VM can be
simultaneously VMotioned to a new host. Limitations are enumerated in the documen-
tation for Storage VMotion.

Figure 9: Storage VMotion

VMware vCenter Host Profiles
VMware vCenter Host Profiles enables more fully automated host provisioning through
the use of host templating. Host Profiles facilitates quickly adding new hardware to the
datacenter by enabling easy application of consistent configurations to the hosts, includ-
ing NFS, firewalls, NTP settings, DNS settings, user profiles, security profiles, and ad-
vanced settings in host configurations. Additionally, Host Profiles provides monitoring
to ensure hosts remain in compliance as configurations change over time.

Figure 10: vCenter Host Proﬁles

vCenter Server Heartbeat
vCenter Server Heartbeat enables monitoring, replication, and rollback for instances of
vCenter Server, enabling effective disaster recovery in the event of loss of the primary
datacenter. It is also particularly useful for protecting vCenter Server when installed on a
physical server.


vCenter Linked Mode
vCenter Linked Mode allows up to 10 vCenter instances (or datacenters) to be aggre-
gated into a single vSphere client, and it includes an advanced search capability designed
to assist in finding VMs throughout the linked vCenter instances. As with the other
components of vSphere, third-party manufacturers can enhance vCenter with SDK and
toolkits.

Figure 11: vCenter Linked Mode

vCenter for Linux ( Technology Preview)
vCenter for Linux is particularly beneficial for organizations not running Windows,
meaning they do not need to purchase and support a single copy of Windows just
to manage their virtual infrastructure. vCenter for Linux is a technology preview, pre-
beta, which can be downloaded for free on the VMware Community Forums: http://
communities.vmware.com/community/beta/vcserver_linux.

VMware vCenter Chargeback
vCenter Chargeback provides the ability for organizations to gauge the cost of providing
and maintaining virtual machines, including rack space, power, cooling, software licens-
es, and associated maintenance. It includes report generation for both internal depart-
ments and external customers, encouraging more efficient use of virtual infrastructure.

VMFS Volume Grow (Dynamic VMFS Expansion)
Dynamic VMFS Expansion is provided by the VMFS Volume Grow mechanism. This
enables the VMFS volume extent to dynamically increase in size along with dynamic
LUN expansion without disrupting the virtual machine.

VMware vStorage Thin Provisioning
VMware vStorage Thin Provisioning allows oversubscription of storage, enabling the
number of allocated “disks” to add up to much more storage than the amount owned.
Advanced management capabilities ensure that over-provisioning does not adversely af-
fect the most important workloads and that it takes advantage of technologies such as
disk duplication. An open API enables leveraging the thin provisioning offered by stor-
age array manufacturers by enabling them to write vSphere plug-ins.


Figure 12: Thin Provisioning

vNetwork Distributed Switch (vDS)
Defined in vCenter Server, the vDS is an abstract representation of multiple hosts defin-
ing the same vSwitch and port groups. This provides the illusion that the VM is con-
nected to the same network as it migrates between multiple hosts. Network management
is simplified by moving the virtual network configuration from the host level to the
datacenter level. The configuration plane and I/O plane have been split, with the I/O
continuing to run on the host level.
This type of switch presents ports to virtual machines and connects to DV Uplink ports
that correspond to physical NIC ports. Policies can be programmed on both DV Port
Groups (VM side) and DV Uplink Port Groups (NIC side).

Figure 13: vNetwork Distributed Switch (vDS) (diagram from Guy Brunsdon of VM-
ware)


Figure 14: vNetwork Distributed Switch Connections (diagram from Guy Brunsdon
of VMware)

VMware High Availability (HA)
VMware High Availability (HA) is a clustering technology that provides virtual machine
high availability on VMware Infrastructure. If one host server fails, all the virtual ma-
chines configured for VMware HA can be restarted on an alternate ESX host. The ESX
hosts must all be members of a VMware HA cluster and must have access to the shared
data stores. Data stores are file systems that store the virtual machine configuration files
and virtual disks, formatted with either VMFS or NFS file system formats.

VMware Fault Tolerance (FT)
VMware Fault Tolerance (FT) provides zero downtime and zero data loss for a virtual ma-
chine by keeping a second instance of the FT-enabled virtual machine in lockstep. If the
host server running an FT-enabled virtual machine fails, a failover to the secondary virtual
machine will take place. The failover is completely transparent to the end-user. The initial
releases of vSphere 4 limit a maximum of one vCPU for a VM running in FT mode.

VMware Distributed Resource Scheduling (DRS)
VMware Distributed Resource Scheduling (DRS) is another ESX clustering technology
that works in conjunction with VMotion to manage load balancing across ESX hosts
while providing a guaranteed quality of service for groups of hosted virtual machines.
When the load on one node in the cluster is unbalanced, DRS uses VMotion to rebal-
ance it, minimizing resource contention and guaranteeing resource levels for virtual
machines. DRS works with VMware HA to ensure that loads are balanced and resource
guarantees are respected in the event of virtual machine redistribution after a host failure.

Figure 15: VMware Distributed Resource Scheduling


VMware Distributed Power Management (DPM)
VMware Distributed Power Management (DPM) optimizes power consumption in a
VMware DRS cluster by consolidating workloads on fewer servers and powering off un-
needed hosts. DRS is an integral part of DPM.

Figure 16: VMware Distributed Power Management

VMware vShield Zones
VMware vShield Zones provides monitoring and control of network traffic based on
corporate security policies defined through the use of vShield virtual appliances. These
policies can be used to align with regulatory compliance guidelines. vShield Zones al-
lows multi-tenancy (internal or external customers on a shared computing resource
pool) while maintaining trust and network segmentation of users and data. This tech-
nology allows monitoring, logging, and blocking of inter-VM traffic within one ESX
host or across multiple ESX hosts. Bridges and firewalls can be created to create multiple
zones based on trust boundaries.

Figure 17: vShield Zones
The management component of this technology is vShield Manager, which is used to
create a firewall for virtual machines. The rules used within this firewall can manage
down to the TCP/UDP port level, operating in Layers 2, 3, and 4 of the network stack.

VMware VMsafe
VMsafe is both an API for security management and control, and a partner program.
VMsafe provides programming interfaces that allow developers to build security solu-
tions, such as antivirus or intrusion detection tools, for the VMware virtual infrastruc-
ture. A virtual machine is protected by these security solutions through inspection of
virtual components, including CPU, memory, network, and storage, meaning protection
at both the host and network layers. By creating tools at this level, there is less resource
load generated within the virtual machine. Similar to how VMware Consolidated Back-


up offloads backup traffic from the VM networks, VMsafe offloads security management
and control from the virtual machines.

VMware Consolidated Backup (VCB)
VMware Consolidated Backup (VCB) provides a mechanism to offload virtual machine
backup traffic from the VM network, the guest OS, and the underlying host by using a
proxy server with direct access to the shared storage. VCB uses virtual machine snapshot
technology, allowing the VCB proxy to back up images of running virtual machines di-
rectly from the storage arrays. A file-level method is provided for daily recovery, such as
for lost files, and a full image method is provided for disaster recovery purposes. The full
image includes the virtual disk(s), configuration file, NVRAM file, log file(s), and any
additional files that represent the state of the machine.

VMware Data Recovery (vDR)
VMware Data Recovery (vDR) offers a disk-based backup and recovery solution that
includes data de-duplication to save on disk storage for backups. It is centrally managed
by VMware vCenter Server. This technology supports restoration of individual files or
entire VM images and automatically monitors VMs that move within the VMware envi-
ronment to ensure backup continuity. It is designed for datacenters requiring backup of
no more than 100 VMs. Full VM backups are taken but allow for individual file restores.

vCenter Orchestrator
vCenter Orchestrator provides a predefined set of workflows that operate with vCenter
and specific features. The version provided with vCenter offers predefined workflows that
can be modified and extended. It allows drag-and-drop workflow creation, using both the
out-of-the-box workflows and access to the vCenter Server API (800+ actions possible).

Figure 18: vCenter Orchestrator—Workﬂow Orchestration

Figure 19: vCenter Orchestrator—Workﬂow Engine


VMware VMDirectPath
VMware VMDirectPath enables connecting the virtual machine directly to a physi-
cal hardware component such as a network card. VMDirectPath enables the virtual
machines to experience native storage and network performance characteristics when
required for particularly demanding workloads.

VMware Record and Replay Virtual Machine Execution
vSphere provides the ability to record and replay the execution of a virtual machine for
forensic or debugging purposes. APIs enable third parties to control this functionality.

The VMware vCenter vCloud Plug-in
The vCenter vCloud Plug-in allows management of, and authentication to, both inter-
nal and external clouds. The vCloud Plug-in uses a variety of techniques to bridge the
VM and cloud and offers a private cloud using resources from both services.

Working with Virtual Appliances
VMware vSphere is a new layer in the traditional infrastructure architecture. In addition
to the required infrastructure hardware and the VMware product suite, many virtual
environments benefit from additional value-added components that extend the native
VMware functionality or provide new features for integrating VMware vSphere with
existing systems and processes. In particular, there are aspects of working with virtualized
applications (Virtual Appliances, OVF, vApp, and ThinApp), which are discussed in the
following section.

Virtual Appliances
Virtual appliances emulate hardware application appliances in that they include a pre-
built and preconfigured application and operating system with simplified management
designed for a specific solution. The operating system and application are packaged
using the industry-standard Open Virtualization Format (OVF) rather than a physical
server. The OVF-formatted virtual appliance can be downloaded and deployed on ESX
or other virtualization platforms. Some virtual appliances emulate traditional physical
hardware appliances such as routers, firewalls, and backup appliances. Others are built
to distribute test or evaluation software or to provide unique functionality such as inline
packet-level patching or storage virtualization.
Whether physical or virtual, the appliance method of application distribution has sev-
eral advantages over distributing software applications to be installed by customers on
standard OS x86/x64 machines. The appliance typically includes a slimmed-down ver-
sion of the operating system tailored to optimally manage the specific application. This
utilizes the computing resources more efficiently while improving reliability, simplifying
troubleshooting, and enhancing security through a standardized application environ-
ment. Appliances generally require less frequent patching. They also eliminate problems
resulting from customers using incompatible hardware or incorrectly installing the ap-
plication.
The downside to hardware appliances is that they are costly, take up rack space, and
require power both to operate and to cool. Hardware appliances result in more under-


utilized servers and also can result in non-standardized hardware in the datacenter. They
have parts that can fail, so duplicate devices might be required to guarantee redundancy.
Additionally, redundant devices might also be required at disaster recovery sites.
Virtual appliances take the application appliance concept to a new level, providing all
of the advantages of hardware appliances and utilizing a specialized operating system in
a controlled environment, but without requiring a dedicated hardware server. Deploy-
ment is simplified, costs are reduced, and high availability is enabled without requiring
duplicate hardware. The virtual appliance can even be replicated off-site, along with the
virtual infrastructure, for disaster recovery without requiring additional appliance licens-
ing or dedicated hardware at the recovery site. Downloading, deploying, evaluating, and
replacing virtual appliances is much quicker and easier.
For these reasons, software manufacturers are increasingly delivering their applications
as virtual appliances. By packaging a database as a virtual appliance, the software manu-
facturer no longer needs to be concerned about what hardware, drivers, and OS version
the application is being installed on. The manufacturer’s best practices are already in-
corporated into the virtual machine, ensuring that it is configured correctly. Complex-
ity is reduced while reliability is improved, and the resource burden on the customer is
also greatly reduced. In addition, the ability to quickly and easily reproduce a problem
through access to the real VM, reproduced in a testing environment at the vendor site,
can help the vendor provide a high support response and customer satisfaction.
Another advantage of virtual appliances is that an organization can collapse more of the
network support services into the virtual infrastructure along with the application serv-
ers. Virtual appliances relying on the network transport, such as firewalls, gain significant
performance advantages by keeping the connections close to the virtual application serv-
ers inside the physical hosts using virtual switches. The virtual appliances will see benefits
at the Layer 2 level, the level at which VMware networking operates. There is less latency
within a virtual infrastructure than routing traffic from the virtual infrastructure out to
the physical network and back. These factors make custom-built virtual software ap-
pliances a natural extension of a virtual infrastructure. As more of the infrastructure is
virtualized, the portability of the entire infrastructure increases, enabling simpler disaster
recovery planning, which makes virtual appliances even more beneficial. The increased
availability of virtual appliances for all types of software will continue to simplify the de-
ployment and support of virtual IT infrastructures.

Open Virtualization Format (OVF)
Open Virtualization Format (OVF) provides an open, secure, portable, efficient, and
extensible format for virtual machines that enables virtualized appliances to be moved
between hypervisors and architectures. The OVF package format provides a complete de-
scription of a single VM or complex multi-VM environments. The package is optimized
for distribution. VM authors can describe “portable hardware” for maximum interoper-
ability or optimum performance. Both compression and package signatures can be ap-
plied. Version 1.0 of OVF has been ratified by the DMTF. For further information on
OVF, see http://www.vmware.com/appliances/learn/ovf.html. VMware vCenter supports
both importing and exporting of virtual machines in an OVF format.


Figure 20: OVF Template creation

VMware vApp
VMware vApp is a logical entity comprising one or more virtual machines. These virtual
machines use OVF, and all components of a multi-tiered application are included in a
package. In addition, operational policies and service levels can be set as attributes for
vApp. The vApp provides a standardized method for describing these operational policies
for the associated application. vApps enable the movement of the application(s) between
internal or external clouds while still supporting the same policies.
The vApp construct was developed to help with the growth of cloud computing by
enabling ease of migration between internal and external clouds and across different
virtualization technologies supporting OVF. vApps provide more flexibility than virtual
appliances, because they allow defining and enforcing policies on an entire application
stack and not just on a single virtual machine. This provides fine-grained control to sup-
port performance aspects of the embedded application(s).
VMware Studio is a tool used to author, configure, deploy, and customize vApps and vir-
tual appliances. VMware Studio is a free virtual appliance downloadable from the VM-
ware site. The tool packages software applications so that they may be easily run within
VMware product platforms and a cloud platform supporting OVF. Both Linux-based
and Windows-based VMs, vApps, and/or virtual appliances can be built with multiple
application tiers included. For further information, see the product description at http://
www.vmware.com/appliances/learn/vmware_studio.html.

VMware ThinApp
VMware ThinApp allows provisioning of applications abstracted from the underlying
operating systems. The application is packaged as a standard EXE or MSI file, including
registry settings or drivers. The application is executed in a sandbox environment. Thin-
App applications can either be stored locally in the operating system or may be placed on
a network share for simplified use and deployment. ThinApp packages reduce complex-
ity and dependencies during the development and update management process. You can
find more information at http://www.vmware.com/products/thinapp/using.html.

3. The Benefits of Infrastructure
Virtualization
The first edition of this book described how virtual infrastructure reduces capital and
operational expenses while improving datacenter agility. While these same benefits
continue to be true, and are indeed still larger today, technological advances enable
organizations to improve performance, management, and security while reducing risk by
virtualizing their datacenters.
VMware vSphere provides the scalability, availability, security, and management
capabilities required to enable complete datacenter virtualization. When combined
with the latest generation of multicore CPUs designed to optimize virtualization
performance, nearly every workload can now run as fast or faster as a virtual machine,
including Microsoft Exchange, SQL Server, large Web servers, and even Oracle
OLTBs. Deploying virtual machines on robust shared storage and encompassing the
enhanced networking capabilities of vSphere should result in overall virtual datacenter
performance superior to that of a physical environment.
Beyond the huge capital and operational expense savings, a completely virtualized
datacenter provides levels of reliability, agility, and recoverability unattainable in the
physical realm at any price. New vSphere capabilities combined with published APIs
enabling plug-ins from leading industry manufacturers enable a virtualized datacenter
that is more secure and better managed than its physical counterpart.
A virtual datacenter is also the platform for implementing robust private and public
cloud strategies, thereby enabling still greater resource utilization efficiency and further
economic benefits.

Capital Expense Reduction
❖ Huge reduction in server hardware expenses resulting from increased server and
storage utilization (the average physical MS Windows server utilization is less
than 15%)
❖ Reduced facilities-related capital costs such as for air conditioners, generators,
PDUs, racks, and cabling
❖ Reduced network and storage switches and cards
❖ Reduced OS licensing costs
❖ Reduced storage costs with VMware vSphere Thin Provisioning
❖ Reduced hardware requirements for high availability and disaster recovery
❖ Reduced cost to purchase PCs, laptops, and remote office infrastructure
❖ Potential elimination of software and hardware requirements for clustering by
using VMware Fault Tolerance

The Benefits of Infrastructure Virtualization / 29

Operational Expense Reduction
❖ Reduced facility costs for rack space and electricity
❖ Reduced procurement costs concomitant to reduced server purchases
❖ Reduced resources required for deployment, replacing, and upgrading physical
servers
❖ Reduced IT staff time spent troubleshooting, patching, capacity planning,
budgeting, and performance monitoring of servers
❖ Reduced downtime expenses resulting from hardware failure, maintenance, or
upgrades and servicing (High Availability, Fault Tolerance, VMotion, Mainte-
nance Mode, and Update Manager)
❖ Reduced network, storage, and disaster recovery management costs
❖ Reduced cost to manage, support, and secure the desktop environment
❖ Reduced power costs of ESX hosts through VMware Distributed Power Man-
agement
❖ Potential reduced cost to manage remote office network infrastructures through
their reduction or elimination entirely

Improved Agility
❖ Ability to provision a private cloud complete with chargeback of IT services to
individual departments
❖ Ability to federate resources between private and public clouds, still further
enhancing both performance and efficiency
❖ Enhanced management, security, and automation capabilities using advanced
VMware features and products such as DRS, Storage VMotion, Update Man-
ager, vCenter Orchestrator, Host Profiles, VMsafe, vShield Zones, VMware
Record and Replay, LifeCycle Manager, and Lab Manager, among others
❖ Increased ability for disaster recovery through VMware encapsulation and
portability as well as advanced features such as VMware Consolidated Backup,
VMware Data Recovery, and VMware Site Recovery Manager (more cost-
effective savings)
❖ Increased service levels from reduced deployment time and cost and from in-
creased responsiveness
❖ The ability to remove legacy hardware while remaining fully operational
❖ Greater service delivery speed through provisioning applications via virtual ap-
pliances
❖ Better support for testing applications and operating systems (via VMware
snapshots and live clones)
❖ Greater alignment of IT with business through increased responsiveness to
business requirements
❖ Ability to enhance organizational “green” initiatives by slashing electricity usage

30 / The Benefits of Infrastructure Virtualization

Reduced Risk
A 2009 webtorials.com IT survey showed server failure as the number one cause of
outage, followed by human error. Virtualization, through VMware High Availability and
Fault Tolerance, enables such fast recovery that server failure is no longer an operational
risk. It also reduces the risk of human error by enabling all testing, patching, and
upgrades to be performed in a virtual sandbox prior to a production rollout. Utilizing
the VMware management, security, and automation products makes a virtualized
datacenter both simpler and less risky to operate than its physical counterpart.

Summary of Benefits
The following is a summary of the benefits that can be achieved through virtualization
of a datacenter.
❖ Reduced risk—Reduces datacenter risk by incorporating superior virtualization
capabilities in performance, management, security, and automation
❖ Server consolidation—Reduces the number of physical servers and associated
costs
❖ Server containment—Prevents server sprawl
❖ Legacy applications—Allows older software applications and operating systems
to run on the latest hardware platforms
❖ Server uptime—Eliminate server failure as an operational concern
❖ Simplified disaster recovery—Encapsulates an operating system, its applica-
tions, and its state within a small set of files, eases recovery, and eliminates typi-
cal plug ’n’ play issues
❖ Standardized hardware for operating system stability—Abstracts the underlying
hardware from the virtual machine, reducing risk and easing underlying hard-
ware changes
❖ Production stability—Simplifies capacity planning and quickens response to
the needs of the business through standardized hardware and encapsulation,
defined roles and responsibilities, and a robust management interface with re-
source trending information
❖ Ease of use—Enhances management and security capabilities
❖ Desktop management—Uses VMware View to ease desktop management, up-
grades, and access
More benefits continue to be found as the technology develops and additional use cases
are identified.

The Business and Operational Case for Virtualization
We described earlier how hardware virtualization enables operating systems
consolidation onto fewer hardware platforms. Several additional benefits are created
by the defining characteristics inherent in virtual machines. In the following sections,
we identify benefits enabled by the following key characteristics of VMware virtual
machines:
❖ Compatibility
❖ Isolation


❖ Encapsulation
❖ Hardware independence

Compatibility
Virtual machine compatibility created by standardization is an essential requirement for
successful VMware implementations. Incompatibility of virtual machines with operating
systems or applications causes instability for some workloads and makes it impossible to
virtualize others. Standardized and compatible virtual machines provide great benefits.
Availability is increased and troubleshooting is simplified when virtual machines
present identical hardware profiles to the operating systems regardless of the underlying
hardware.

Isolation
The isolation characteristic of virtual machines is critical in production and software
development environments. Virtual machine isolation has significant availability and
security benefits. This feature allows users to safely run multiple virtual machines on
one host. Security-conscious organizations need confidence that information from one
virtual machine cannot leak into another and that a compromised virtual machine
cannot have access to the memory, data, or network traffic of any other virtual machine
on the host. The isolation property of virtual machines ensures that this does not
happen.
The availability and stability benefits of isolation are necessary because in a production
or software testing environment you cannot afford to have one virtual machine disturb
the operation of other virtual machines. In a software development environment,
test programs may crash or misbehave, but this does not affect the operation of other
environments that share the same computer.

Encapsulation
Encapsulation makes virtual machines portable and easy to manage by representing
physical hardware in a set of files on disk. A virtual machine is backed up by copying the
set of files. Of course, a backup program can also be run within the virtual machine to
perform incremental or full backups.
As simple as this seems, this is a major benefit. Having a whole operating system
instance in a set of files allows for virtual machine portability. A system administrator
does not have to worry that a particular configuration is composed of an operating
system install, set of updates and patches, a set of registry settings, and appropriate
documents and settings directory files, all configured for specific hardware. A virtual
machine can be replicated, copied to portable media, or archived just like any standard
data files, then executed on any supported virtualization platform at a later time. This
characteristic, along with hardware independence, provides one of the greatest benefits
of a virtual infrastructure: the virtual machine mobility and portability that enable
simplified disaster recovery and hardware migrations.


Hardware Independence
When coupled with the properties of encapsulation and compatibility, hardware
independence provides the freedom to move a virtual machine from one type of x86
computer to another of the same instruction-set family without making any changes
to the device drivers, operating system, or applications. Hardware independence also
means that you can run a heterogeneous mixture of operating systems and applications
on a single physical computer. In your datacenter you can mix and match servers from
multiple vendors with ease, and also use a single group of servers to run both Windows
and Linux without any alteration or reinstallation of operating systems. However, you
might want to standardize on a few hardware vendors for other reasons, such as spare
parts, standardized hardware configuration, and simplified purchasing and discounts.
The flexibility of changing vendors using the layer of virtualization provides for a form
of freedom that has seldom been seen before.

Return on Investment (ROI)
It is hard to imagine a more clear and compelling ROI than is obtainable from
implementing a VMware Virtual Infrastructure. The primary areas of savings result from
virtualizing production datacenters, backup and disaster recovery, and desktop systems.

Production Datacenters
The most obvious savings result from consolidating servers. While consolidation ratios
can vary widely depending on the type of servers being virtualized, VMware DRS
enables effective load balancing of workloads among multiple ESX servers. It is not
uncommon to achieve average consolidation ratios of 20 or more virtual machines per
two-CPU late-generation multicore servers. An organization virtualizing its production
environment, even allowing for redundant ESX hosts and resource intensive tier 1
applications, may be able to consolidate 85% or more of its physical servers, along with
their associated power consumption, cooling costs, and maintenance expenses. The
demand for less power can lead to further savings from reduced requirements for air
conditioners, PDUs (Power Distribution Units), UPS (Uninterruptible Power Supply)
devices, and generators. Slashing the number of servers similarly slashes expenses for
rack space and other maintenance-related costs. This can be particularly significant
for organizations running out of datacenter space or using outsourced datacenters that
charge by the square foot or rack.

Storage
Virtual storage becomes a reality in a VMware virtual infrastructure. Features that SAN
manufacturers sometimes require to be purchased separately, such as server snapshots,
migrations, and multipath software, are provided through VMware’s disk abstraction.
The thin provisioning capability of vSphere provides a higher level of utilization.
Resource consolidation provided by ESX allows all virtual servers to benefit from high-
performance SAN storage without incurring the per-host connectivity fees traditionally
associated with network storage. Purchasing and maintenance costs for host bus adaptors
(HBAs) and storage switches can be greatly reduced because multiple virtual machines
of all types and priorities can be connected to high-end shared storage through storage


virtualization. Also, because LUNs can span many virtual machines, managing the
storage environment is easier as well.

Network
VMware’s virtual switches provide the ability to seamlessly extend the physical network
into the virtual network. Network virtualization expands the network infrastructure
without requiring additional switch ports. Consolidating 100 servers on six ESX hosts
can result in saving 60 network switch ports while providing multiple link redundancy
for each virtual server. This same configuration can provide SAN connectivity to the 100
virtual servers while avoiding the cost of 172 FC switch and HBA ports.
Additional virtual switch features such as 802.1q (VLAN tagging) and 802.1ad (port
trunking or link aggregation) provide increased flexibility and functionality to the
virtual and physical network infrastructure. The VMware vNetwork Distributed Switch
(vDS), included with VMware vSphere Enterprise Plus, enables application of policy-
based network capabilities that move with a virtual machine during live migration. This
allows virtualization of more servers while still meeting security and regulatory policies.
Consolidation of the virtual machine management reduces the amount of time spent on
virtual network administration.

Management
The ability to manage all VMware vSphere components from a single point of view
(pane of glass) makes virtual infrastructure administration less complex than physical.
An IDC and VMware study showed the ability for administrators to effectively
manage over three times more servers in a virtualized environment than in a physical
environment.

Software Licensing
Savings frequently result from operating system licensing policies in a virtual
environment. For example, licensing Microsoft Windows Server Datacenter Edition for
the underlying CPUs of the ESX host enables an organization to run unlimited instances
of Windows Server on the host. The same is true for SQL Server Enterprise and for
BizTalk, as well as other applications from various vendors.

Backups and Disaster Recovery
Virtualization enables backups to be performed at the virtual disk (*.vmdk) level as well
as standard file and block-level backups. This results in less effort and money spent by
organizations trying to meet shrinking backup availability windows.
Disaster recovery in a virtual environment is not only functionally much more effective,
it is more cost-effective as well. Far fewer servers are required at the recovery facility, and
they do not need to be the same brand or type of servers as the production servers. Both
data and virtual machines can be continuously replicated to the recovery site, enabling
inexpensive recovery of a VMware virtual infrastructure. Disaster recovery testing can
also be done much less expensively and without requiring personnel to go on-site at the
recovery site facility. VMware’s Site Recovery Manager (SRM) automates the disaster


recovery process, enabling still further increased effectiveness with reduced staffing
requirements.

Desktops
Virtualizing the desktop with VMware View (VDI) enables another realm of savings
and very positive ROI. These savings result from reducing the frequency of PC and
laptop upgrades and from reduced power and operational costs when using thin-client
or zero-client terminals or lower-cost PCs or laptops. Maintenance costs are lowered and
administrative requirements are reduced. An IDC analysis (Michael Rose and Randy
Perry, September 2009) concluded that organizations deploying VMware View saved
on average over $610 per supported end-user per year compared with organizations
using unmanaged PCs, with an additional $122 per year when using the capabilities
of VMware View Premier. The IDC white paper can be found by registering for the
desktop virtualization kit on www.vmware.com. An example of a desktop ROI can be
found on http://www.bythebell.com/2009/09/the-desktops-may-be-virtual-but-the-roi-
is-real.html.

ROI Case Study: A Community Bank
The following example of ROI savings is based on a VMware deployment by a major
San Francisco Bay Area community bank. This bank consolidated 102 of its 105 servers
onto six ESX hosts. The bank realized a cost savings of $1.6 million over a five-year
period versus an initial investment of $270K, resulting in an investment payback period
of only 10 months.
While the majority of savings result from server consolidation and containment,
substantial savings also result from reducing DR costs and electricity requirements,
shown in the figures on the following page.

ROI/TCO Calculator
The VMware ROI/TCO Calculator provides a method for determining ROI and TCO
(total cost of ownership) based on server consolidation scenarios. It provides support for
measuring the savings or estimating potential savings when deploying any combination
of VMware vSphere, VMware Lab Manager, and VMware View.
The tool focuses on comparing the costs of an existing physical computing environment
against the same environment after being converted into a VMware virtual infrastructure
environment. The tool asks five to ten questions about the existing environment to
determine the cost to deploy and maintain the infrastructure. The questions are about
industry, business, and technology drivers; location; current assets, including the number
of servers to be virtualized; and time required for a typical new server deployment,
among others.
This information is used to set default values for 200 additional metrics used to
determine the cost to deploy and maintain the existing and future virtualized
infrastructure. These values can be modified to better represent your organization or to
create what-if scenarios.
Initially, the calculator quantifies the cost of maintaining the existing physical


infrastructure. Then it calculates the costs for converting the environment into a
VMware virtual infrastructure. The cost savings are then quantified for management
overhead, productivity improvements, risk reduction, and other technical and business
benefits. The ROI/TCO Calculator produces final outputs, including the cost
investment, savings, and additional benefits to set up, deploy, and support the new
virtualized infrastructure.

Figure 21: Physical versus virtual DR costs

Figure 22: Physical versus virtual electric costs
Results of the analysis can be viewed comprehensively or individually for VMware
vSphere 4, Lab Manager, and VMware View deployments. The calculator results can
be used to analyze different scenarios by adjusting variables and saving multiple result
sets for comparison. Results can be reviewed online or in export PDF, RTF, or raw
data (XLS) formats. The VMware ROI/TCO Calculator is a valuable tool if you are
considering or trying to justify a large-scale server, desktop, or lab virtualization project.
The calculator is also an excellent tool for verifying ongoing savings with accurate values
after a deployment.
The online ROI/TCO calculator can be found at http://www.vmware.com/products/vi/
calculator.html.

4. Use Cases for Virtualization

VMware is used by 96% of the Fortune 1000 in a production capacity, but the aver-
age virtualization penetration within a datacenter is only around 30%. While effective
as a point solution, the full benefits of virtualization are realized when it is deployed
as an encompassing platform for the entire datacenter, incorporating both storage and
network along with servers. VMware vSphere provides the scalability, availability, man-
agement, and security requirements enabling a 100% virtualized datacenter, which in
turn provides the foundation for both effective cloud computing and service-oriented
architecture.

Production Environments
A virtualized datacenter reduces risk by providing high availability for any server, in-
dependent of traditional clustering technologies like MSCS and VCS. It enables more
effective disaster recovery by being able to continuously replicate all servers off-site. It
maximizes reductions in equipment and facility expenses while simplifying administra-
tion.

Cloud Computing
VMware vSphere transforms a datacenter into an internal cloud by enabling a pool
of virtualized resources within a datacenter. It also enables seamless resource shifting
between internal and external clouds.

Service-Oriented Architecture
A virtualized datacenter is a key enabler of service-oriented architecture (SOA) by au-
tomatically provisioning IT infrastructure in response to application requirements. Dy-
namically allocated virtualization pool resources optimize service requests. Additionally,
as applications increasingly are written to communicate directly with the ESX hypervi-
sor, the requirement for a traditional full operating system layer with its inherent greater
costs and inefficiencies is eliminated. The better the integration with the hypervisors, the
lower the total overhead.

Software Test/Development and Testing
ESX got its start in the test/development area. Because ESX enables administrators to
easily make a snapshot copy of a server, the snapped copy can then be patched, up-
graded, or migrated in a separate virtual environment that uses VLANs to emulate the
production network. Also, VMware Lab Manager takes software development to a new


level by enabling developers to provision multiple sets of servers in seconds, as well as
the ability to share and archive environments for QA testing or troubleshooting.

Disaster Recovery
Research organizations such as Meta and Gartner emphasize the high probability of
business failure if an organization suffers a disaster such as a fire or flood. This results
from the lack of an effective disaster recovery plan or the failure of a disaster recovery
plan. Disaster recovery plans for traditional physical computing infrastructures are very
complex and expensive to implement and maintain, usually requiring a mirror of the
production infrastructure at a remote site. These plans are difficult to test effectively
before they are needed and they tend not to work well, if they work at all. Conversely,
VMware Site Recovery Manager enables disaster recovery that is significantly less com-
plex, more affordable, and testable. All virtual machines, not just a small subset deemed
as mission-critical, can be included in a VMware Infrastructure disaster recovery solu-
tion. The virtual machines with their networking and storage can be continuously repli-
cated from the production datacenter to the recovery facility where they are ready to be
activated with a single click. Virtual desktops and virtualized applications can also be
replicated, enabling users to connect to their applications and data from anywhere they
can access the Internet using a browser.

Remote Offices
Small, remote offices frequently can be inexpensively virtualized using VMware vSphere
Essentials, which is a lower-cost version of vSphere designed for smaller deployments,
making virtualization in small offices affordable while providing the encapsulation and
mobility features that simplify backup and disaster recovery from remote offices.

Desktops
VMware View extends virtualization to the client by running desktops on centralized
vSphere hosts. This computing model gives employees far more flexibility by being able
to work from any PC, Mac, Microsoft Windows terminal, or zero-client device. When
employees log on, they are directed to their virtual desktop, applications, and data. Un-
like server-based computing that relies on a shared Windows kernel to achieve multi-
user access, VDI truly provides users with their own customized desktops while reducing
the cost of hardware and simplifying administration.

5. Designing a Virtual Infrastructure

As you virtualize your IT infrastructure, you need to address the planning, design,
implementation, management, and technical details of VMware vSphere. The VMware
Server Consolidation Methodology provides a proven approach to virtual infrastructure.

VMware Server Consolidation Methodology
The VMware Server Consolidation Methodology consists of several phases designed to
support a smooth transition from your existing environment to a virtualized infrastruc-
ture. Small tactical deployments of VMware vSphere can be done without using the full
methodology, but skipping any of these phases for a larger infrastructure deployment
usually leads to a more expensive and under-optimized infrastructure. All of the phases
are recommended and are typically used to complete a full end-to-end virtualization
strategy.

Phase 1—Assessment and Planning Phase
The assessment and planning phase includes conducting a virtualization assessment to
identify virtualization candidates and their resource requirements. Application require-
ments and interdependencies are also determined. During this phase, inventory and per-
formance data is collected to develop implementation options. A gap analysis provides
detailed information about missing components, staffing, and operational procedures.
Phase 1 can include training, identification of options for the design, and a proof-of-
concept implementation for demonstration purposes.

Phase 2—Design Phase
In the design phase, customized tactical implementation details, technical blueprints,
management and assembly guides, and test plans are created for deployment. These
are used for the deployment of VMware vSphere during the third phase. Phase 2 can
include a pilot program to test aspects of the design to validate alignment with business
requirements.

Phase 3—Build/Implementation Phase
Deployment and testing occur during the build and implementation phase. Virtual ma-
chines are built or imported using VMware Converter or other physical-to-virtual (P2V)
migration tools. Both unit-level testing of individual components and testing of multiple
system components working together are conducted to ensure that new virtual machines
fit correctly within their allocated resources and environment. User acceptance tests are
conducted at the end of this phase.

Designing a Virtual Infrastructure / 39

Phase 4—Management Phase
The management phase is when maintenance and ongoing refinement to the infrastruc-
ture occur. Performance is analyzed and tuning is performed to ensure compliance with
service level agreements. Service level agreements can cover availability, performance,
and other areas essential to your business requirements. A check is typically performed
once a year by internal teams or external consultants to identify deviations from current
VMware best practices and supported configurations. Over time, best practices evolve
based on technology updates and feedback from users.

Identifying Virtualization Candidates
When beginning a server virtualization project, the first question is often which servers
can or should be virtualized. Large-scale server virtualization projects done without accu-
rate visibility into the existing infrastructure can undermine predictability and efficiency
and increase risk. A systematic approach to identifying virtualization candidates and
computing capacity is required to realize the maximum benefits from a server virtualiza-
tion and consolidation project.
Almost all operating systems and applications are potential candidates for virtualization.
However, some may require analysis to determine the resources required to maintain
service levels. When determining VMware vSphere resource requirements, Virtualization
Assessments consider many factors, including utilization of the original physical ma-
chine and its hardware, operating system, and applications.
Good candidates can maintain or improve performance levels in a virtual machine post-
virtualization. The following are items that need to be considered or are typical qualifiers
when determining virtualization candidates.

Vendor Support
Does the vendor support their application running within a virtual machine?
❖ If yes, then proceed with migration.
❖ If no, then follow up with vendor to determine plans for support.
❖ If the vendor has no plans, determine whether using virtual to physical (V2P)
migration during troubleshooting will satisfy internal and vendor support
needs. This allows reproducing a problem on a physical platform.

Resource Requirements
❖ If more than eight CPUs are required to maintain service levels, the server may
not be a candidate for virtualization.
❖ If USB ports are required, a USB-over-IP hub might be considered to provide
the resource. Otherwise, VMware Server, which provides greater hardware
compatibility, may be an alternative.

Real-time Data Acquisition
❖ If the application and associated devices providing data to the application can
ensure no data loss, then the virtual machine can typically be virtualized. This
typically requires a cache mechanism at the device site.

40 / Designing a Virtual Infrastructure

❖ If not, then the application stays physical.

Conducting a Virtualization Assessment
There are several phases in a comprehensive virtualization assessment. First, you must
inventory the hardware and software components of a system. The second phase in-
volves analyzing application resource utilization for the target physical servers over the
course of several weeks. During the final phase of the assessment, the resource utilization
of the virtualization candidates is analyzed to determine the aggregate CPU, memory,
network, and disk I/O requirements. This enables proper sizing of the target VMware
Infrastructure.
Manually gathering inventory data from a large number of servers can be time-consum-
ing and difficult. Some tools that can be used to help with assessments include VMware
Capacity Planner, IBM CDAT, Microsoft Msginfo32, Microsoft Srvinfo, the UNIX
System Activity Reporter (SAR) tool, and Platespin PowerRecon. These tools provide
support for one or more operating systems. VMware Capacity Planner is used during
a VMware Virtualization Assessment to gather an inventory and performance metrics.
The typical data collection requires 30 days to account for any workload variance in the
month. Afterwards, optimization scenarios are run against the target hardware platform
to produce consolidation ratios.

Inventory
The first part of the virtualization assessment is the gathering of inventory data for the
systems that are to be virtualized. Specific items to consider include:
❖ Type, speed, number of CPUs, and number of cores per CPU
❖ CPU average and peak usage
❖ RAM available and peak usage
❖ Disk space available and used
❖ Network I/O used
❖ Type and number of NICs
❖ Operating system and version (include patches and service packs)
❖ Applications installed
❖ Special hardware used, which determines whether or not it can be virtualized
on ESX

Application Resource Considerations
Application resource utilization must be considered and analyzed. Analyzing the re-
source utilization of each operating system and application pairing over a 30-day busi-
ness cycle makes the expected resource load evident. Interactions between the existing
resource utilization and additional overhead caused by virtualization must be considered.
CPU virtualization incurs the least amount of overhead, but disk I/O increases CPU
utilization and context switching.
The subtle interactions between these various components, combined with translating
individual application performance to a consolidated platform on a new hardware clus-
ter, makes virtualization planning a complex task.


Tracking resource utilization over a one- to three-month period provides important data
to determine candidates for virtualization. The components to measure include CPU,
RAM, disk I/O, and network I/O. Average and peak usage, base characteristics of de-
vices, and the target hardware platforms influence the migration and deployment design.
To manually conduct this analysis, a programmer could use statistics gathered using
the Perfmon for Microsoft Windows operating systems or the System Activity Reporter
(SAR) tool for UNIX or Linux operating systems, but these tools cannot translate the
data to useful virtualization scenarios. Using a capacity analysis tool such as VMware Ca-
pacity Planner provides a more accurate model and greatly simplifies the consolidation
planning process.

cpu
ESX 4 supports a maximum of 64 cores on its host platform, and up to eight vCPUs in
a virtual machine. It is recommended not to provision a virtual machine with more vir-
tual CPUs than the host has CPU cores.
Additional metrics which influence the target virtualization architecture include CPU
queue depth, %READY (the percentage of time an instruction was ready to execute
but could not run due to lack of CPU resources), user and system time, as well as other
items. The %READY measure is valid for both physical and virtual machines.
When sizing the target host hardware, it is important to anticipate the number of
vCPUs required for the virtual machines. It is a best practice to initially provision a
single vCPU per virtual machine. After monitoring performance and utilization, addi-
tional vCPUs can be added if necessary. Multi-threaded applications often benefit from
multiple vCPUs.
An estimate of four vCPUs per core can be used for capacity planning. Actual consolida-
tion ratios vary based on performance data and target ESX hardware platform. When
selecting CPUs for host systems, larger cache size can provide performance gains. Large
cache sizes (both level 2 and level 3 caches) typically yield higher performance gains.
Virtualizing CPU resources creates lower overhead than memory, disk, or network I/O
virtualization. CPU virtualization adds varying amounts of overhead, depending on how
much of the virtual machine workload can be run in direct execution.

ram
VMware ESX 4.0 supports a maximum of 1TB of RAM on its host platform. Each
guest OS can be allocated a maximum of 255GB of RAM (guest OS limitations apply).
Faster memory makes a significant difference to application performance as consolida-
tion ratios are increased.
All operating systems use page tables to map virtual memory to physical memory. The
VMware hypervisor allocates physical memory to the guest OS using shadow page
tables. It then uses a separate table to map those allocations to physical memory. This
process allows memory sharing between guest operating systems and can reduce memory
latency. We also use CPU techniques such as Intel’s Extended Page Tables™ (EPT)and
AMD Rapid Virtualization Indexing (RVI). Both of these provide support for Memory
Management Unit (MMU) virtualization increasing performance, particularly for

42 / Designing a Virtual Infrastructure

MMU-intensive workloads. Some benefits include higher throughput and lower CPU
utilization.

disk i/o
VMware ESX 4.0 can support up to 64TB (minus 512B overhead) of storage per host
using a file block size of 256MB. Each guest OS can be allocated a maximum of 2TB
(minus 512B overhead) of storage.
The analysis of disk I/O is a critical part of resource planning for a server consolidation
project. Disk access can sometimes be the constraining resource for application perfor-
mance in both physical and virtual computing platforms. In a VMware Infrastructure,
storage resources cannot be dynamically migrated for performance optimization or have
the quality of service features that CPU and memory provide. Storage VMotion can be
utilized to manually move virtual disks to assist in optimizing disk performance. Disk
I/O is the virtualization resource that requires the most overhead, consuming CPU time
and creating additional interrupts.
Depending on throughput requirements, application servers can rely heavily on disk I/O
for performance. It is best to map SAN LUNs to application servers according to their
required performance levels. A file server may be optimally configured for sequential ac-
cess utilizing a RAID 5 LUN built on 7200 RPM Fibre Channel or SATA disks. A typi-
cal LUN for high-performance applications such as databases would be built on RAID
10 using many 15,000 RPM disks for optimal random access speeds. High disk I/O
performance can come at a price.
When using iSCSI, TCP/IP can consume large amounts of CPU cycles unless paired
with an iSCSI TOE (TCP Offload Engine) card. Also, Fibre Channel requires overhead,
as the host has to use CPU cycles to compute disk I/O. However, with current hardware
and improvements in the vSphere iSCSI software initiation, CPU overhead should not
be a concern.
vSphere provides support for NFS Jumbo Frames when using 1Gbps and 10Gbps NICs.
NAS, like iSCSI, can benefit from TOE cards. Some vendors claim performance gains
up to 30% depending on load.
Measuring the anticipated disk I/O profile of the servers to be consolidated is essential
for accurate disk I/O planning and maintenance of service levels. A virtualization assess-
ment provides the level of detail necessary to plan for thoughtful I/O consolidation.
In the case of large Exchange servers or database servers, performance considerations for
I/O may be important. Without the benefit of a virtualization assessment, users often
create a LUN and continue to add virtual machines to it until performance degrades or
the capacity is exhausted. To recover performance on the oversubscribed LUN, virtual
machines need to be migrated off the congested LUN, costing additional administrative
time and SAN overhead. This approach rarely leads to optimal performance and capac-
ity utilization. Support for N_Port ID Virtualization (NPIV) has been added to allow
each virtual machine to have its own World Wide Name (WWN), allowing for VM-spe-
cific QoS. Additionally, full support for round-robin HBA load balancing is included, as
well as support for third-party multipathing plug-ins like EMC’s Powerpath/VE.
When sizing VMFS storage, measure disk throughput in inputs/outputs per second
(IOPS) for each candidate server collected during the virtualization assessment. A good


rule of thumb is that a Fibre Channel disk can support approximately 100 IOPS RAID
groups. LUNs should be built with enough disks to support the anticipated IOPS for
the candidate servers. This planned approach can lead to predictable performance with
fewer migrations to reallocate virtual machines from oversubscribed LUNS.

network i/o
VMware vSphere 4.0 can support Fast Ethernet (100 Mbps), GigE (1000 Mbps), and
10G (10,000 Mbps, with support for TCP Segmentation Offload (TSO) and jumbo
frames.
Virtualized network I/O on ESX hosts, like disk I/O, creates greater overhead than CPU
or memory virtualization while simultaneously adding to the CPU load for each I/O.
Much of network performance (like disk performance) is external to the ESX host and
out of the control of the VMkernel and the Virtual Machine Monitor (VMM). These
factors should be considered when identifying candidates for server virtualization.
These additional layers of overhead need to be taken into account when sizing host serv-
ers and considering which systems to use for virtualization. Several improvements in
networking efficiency have been introduced that reduce VMM overhead and decrease
network I/O-induced CPU utilization. The latest version of the VMXNET virtual net-
work driver includes TCP Segmentation Offload (TSO) support which allows previously
CPU-intensive TCP segmentation operations to be off-loaded to the physical network
adapter. Jumbo Ethernet frames (frames with a size greater than 1500 bytes) are also
supported; they greatly reduce the number of packets needed to transmit data. Each I/O
operation removed from the CPU workload through TSO and jumbo frame support
reduces virtualization overhead and increases available CPU time for virtual machine
execution.
When considering a server’s workload for virtualization, it is important to consider the
VMM and VMkernel overhead created by network I/O, and whether TSO or jumbo
frames can be used to reduce this overhead.

Agile Approach to Designing the Virtual Datacenter
Designing a virtual datacenter requires significant consideration for integration with
hardware, software, policies, and procedures within a business. A design includes more
than configuration of vCenter and ESX/ESXi. In reality it starts with the underlying
requirements and constraints from the business.
In many situations, a waterfall design approach does not always fit due to time, budget,
or staffing constraints. An agile approach provides an opportunity for starting a design
while the planning phase is underway, and for starting the implementation during the
design phase. A waterfall approach is sequential while an agile approach supports paral-
lelism in the process.
As part of the design process, factor in the justification and impact of each design choice.
This information provides a clear picture during the review process with different busi-
ness units and the core IT teams. It also ensures better integration with other VMware
and third-party technologies.

6. Building a VMware vSphere
Environment
Selection and configuration of hardware components are critical to the performance,
scalability, and reliability of a VMware infrastructure. The details of the host hardware
don’t matter to the guest operating systems in a virtual infrastructure, but the underlying
hardware architecture and configuration have considerable effect on the performance,
scalability, and reliability of the VMware vSphere clusters.
The VMware vSphere hardware cluster consists of two main components: the host server
hardware and the storage network and its subsystems. Two additional hardware compo-
nents, the data network and the backup infrastructure, play an important part, but are
usually external to the virtual infrastructure. The most important requirement for system
stability and VMware support of ESX or storage hardware is compliance with the VM-
ware Hardware Compatibility List (HCL). The tight coupling between the hypervisor
and hardware depends on using tested and compatible hardware. The VMware HCL can
be found on the VMware Web site: http://www.vmware.com/resources/compatibility/.

Server Hardware
In a virtual infrastructure, the function of the host server hardware is to provide a con-
tainer (processing, memory, and I/O) for guest operating system execution. The nature
of consolidated resource sharing on an ESX host or cluster makes selection and optimi-
zation of its underlying hardware critical to efficiency and performance. The ROI and
success of a VMware vSphere project depend largely on hardware selection.
When selecting ESX host hardware, CPU core density, memory, host server mother-
board, I/O, and scalability must be considered for maximum cost-efficiency, scalability,
and performance.

CPU Counts and Core Density
As CPU speed and core density increase, host servers are able to provide processing pow-
er for more guest operating systems. Multiple multicore processors provide low latency
performance during the simultaneous execution of multiple guest operating systems
within ESX. A typical processor configuration consists of at least two quad-core 64-bit
compatible processors with Intel VT or AMD-V virtualization extensions for maximum
compatibility and performance. High CPU core densities and large processor caches
provide greater ESX scalability and performance than increased CPU MHz alone.
The high core counts in today’s CPUs allow for high consolidation ratios on servers with
only two physical processors. Four or more physical processors are recommended if the

Building a VMware vSphere Environment / 45

host will be running virtual machines with quad-processor SMP. To run ESX 4.0, 64-bit
processors are required.

Memory
Memory is a critical component of virtual infrastructure hardware. Today’s latest gen-
eration of 64-bit-capable servers feature large memory capacity and high-speed front
side buses. The high consolidation ratios supported by servers with high processor core
counts require servers with large amounts of memory to fully utilize the available proces-
sor power. Even with ESX memory sharing and memory reclaiming capabilities, ESX
needs additional memory available to satisfy Distributed Resource Scheduling and High
Availability services.
It is not uncommon for ESX to reach memory capacity while still having unused proces-
sor power available, making memory the limiting resource to scalability. A dual-proces-
sor eight-core server hosting 20 to 25 guest operating systems can often utilize 32–48GB
of RAM when accounting for Distributed Resource Scheduling and overhead.

Host Server Motherboard Considerations
The continuing increase in processor speeds only translates to incremental increases in
overall system speeds without fully optimized subsystems to support data movement
within the host server. The latest server generation features up to 1600 MHz front side
bus speeds to support the newest PCIe and memory speeds and processor core densities.
ESX benefits from parallel processing, increasing guest operating system scalability with
the addition of cores and processors. However, it is the internal bus, memory, peripheral
speeds, and system cache that have the greatest effect on the application and guest oper-
ating system speeds.

PCIe Slots
Due to the scalability enabled by quad-core processors and 64-bit memory addressing,
it is desirable to configure servers with a large number of PCIe slots for network and
storage cards. PCIe buses enable dedicated I/O paths, allowing for linear performance
scalability. It is not unusual for 8- or 16-core ESX hosts to use six or more network in-
terfaces and at least two storage I/O cards. The maximum number of supported cards is
provided in the “Configuration Maximums” paper on the VMware Web site at http://
www.vmware.com/pdf/vsphere4/r40/vsp_40_config_max.pdf.
Processing virtualized storage and network I/O can consume a large number of CPU
cycles, creating context switches and a considerable amount of FSB (front side bus)
traffic. In a physical system, data can be moved across the system buses five times after
entering the server before it gets to the CPU. Processing 1GB/s of network traffic can
consume up to a GHz of CPU cycles. This situation is exacerbated by time-sharing be-
tween a large number of guest operating systems on an ESX host. This results in a large
percentage of host server CPU power dedicated to processing network and storage I/O,
ultimately limiting system scalability. ESX supports TCP segmentation offload, jumbo
frames, and the use of TCP Offload Engine (TOE) network cards. These NICs offload

46 / Building a VMware vSphere Environment

TCP processing from the CPU, reducing I/O overhead and freeing CPU cycles. New
technologies like Infiniband connectivity overcome these limitations by using remote
direct memory access and low latency 10GB/s consolidated into a single link (two links
for redundancy). However, with today’s processors, you will need to keep in mind that
investing in additional processors or processor speed might be more cost-effective than,
for instance, buying iSCSI Initiators.

Storage Hardware
Storage networks and shared storage arrays are critical to a well-architected virtual infra-
structure. Storage design, configuration, and performance are the most important and
complex hardware subsystem considerations in virtual infrastructures. Virtual machines
are stored in raw LUNs, NFS shares, or VMFS file systems and are executed on any of
the servers in a VMware vSphere cluster. Networked storage is necessary for all of the
advanced VMware vSphere management features, including VMotion, HA, FT, DRS,
vDR, and VCB. Storage-related issues are the most common and problematic VMware
vSphere support issues.
The full potential and savings of virtualization can only be realized when the storage
system is optimized and managed as a single functional unit within the virtual environ-
ment. There are a wide variety of storage arrays tested and approved for compatibility
with VMware vSphere 4, from inexpensive cluster solutions and mid-range arrays to
large enterprise storage systems. Across this range of arrays, feature sets and performance
vary widely. All storage arrays on the VMware HCL can be integrated into a VMware
vSphere environment, but there are some storage system characteristics more optimized
for VMware vSphere. Several storage vendors offer VMware vCenter plug-ins to enable
monitoring and managing the virtual environment from a single pane of glass.
VMware vCenter Server provides interfaces for dynamically managing all aspects of a
VMware vSphere environment, including hosts, clusters, guest operating systems, and
virtual networks. Storage and VMware vSphere environments are often managed by dif-
ferent teams, with storage systems generally being less agile and dynamic than the rest
of the VMware virtual infrastructure. Storage as an extension of a VMware virtual infra-
structure must be able to quickly react to change if it is to add to, and not detract from,
the overall value of the VMware virtual infrastructure. The ideal situation is to have the
storage system managed as a part of the VMware virtual infrastructure, extending the
boundary of the virtual infrastructure to include the physical and logical storage. Newer
storage systems that provide storage virtualization features allow a more dynamic and
manageable storage layer for ESX clusters.
Infrastructure performance issues that are not caused by oversubscribing the server’s
capacity are most often traced back to storage architecture. Server processor speeds have
increased approximately 12 times faster than storage performance has, fostering the
perception that simply throwing more or faster processors at a performance issue will
cure it. One vital but often underappreciated requirement for VMware vSphere storage
systems is the need for linear performance per capacity scaling.
Virtual infrastructures have a very different storage profile from those of most traditional
workloads. Capacity expansion is driven by adding virtual servers, often with higher-


load virtual machines being added in later phases of the projects. Maintaining multiple
virtual disk files in each LUN combined with the resource time-sharing inherent in
virtualization results in increased random disk I/O with each virtual server added to the
infrastructure. A storage system selected for a VMware vSphere environment should be
able to scale performance linearly with virtual server growth to the upper limits of its
capacity growth.
Managing quality of service is essential when consolidating several different workloads
with various service level needs on a VMware vSphere cluster. VMware vCenter Server
supplies robust quality of service features to maintain desired service levels across differ-
ent priority workloads running on shared hardware. The Distributed Resource Scheduler
provides effective grouped and granular controls over server CPU and memory resource
shares and limits; however, there is no effective way to manage disk access performance
levels from within the standard VMware vSphere management systems today.
To maintain true quality of service within a VMware Infrastructure, the storage systems
should be able to easily differentiate and manage LUN performance for different virtual
server workloads. VMware infrastructure administrators should be able to manage stor-
age quality of service as easily and non-disruptively as CPU and memory QoS adminis-
tration.

ESX
ESX is the primary datacenter virtualization platform within VMware Infrastructure.
ESX is a thin layer of software that runs on bare-metal x86/x64 hardware. It abstracts
available hardware resources, such as processor, memory, storage, and networking, and
multiplexes them among virtual machines running unmodified commodity operating
systems.

Figure 23: vSphere Client view of an ESX Server


Prerequisites
ESX can be deployed on a variety of server hardware platforms. The exact configura-
tion of server hardware that is certified to run ESX is frequently updated. The VMware
Compatibility Guide contains the most current list of systems certified to run ESX
(http://www.vmware.com/resources/compatibility).
VMware ESX 4.0 will only install on a server with 64-bit x86 CPUs.

Installation
The high-level steps for installing ESX are:
1. Collect information about the system on which ESX is to be installed.
2. Decide on an installation method from the following options:
a. Local CD-ROM or DVD-ROM drive.
b. A remote CD/DVD drive using the remote system software provided
by the manufacturer of the hardware system.
c. Automatic Kickstart-based scripted installation. (See the section
Kickstart Scripted Installation, below.)
d. Download ESX software ISO image and burn a CD or DVD.
3. Choose and identify the correct network interface, as the ESX Service
Console requires this. By default, ESX uses the first interface discovered
(vmnic0) as the Service Console interface. On a system with multiple
network interfaces, it is important to know which NIC shows up as
vmnic0. Viewing the BIOS and identifying the MAC address of the NIC
can determine this.
4. Install ESX.

Local CD or DVD Drive Installation
After the target server is booted from an ESX install CD, two options are available for
continuing the installation:
❖ A mouse-based graphical installation is available. This is the easiest and recom-
mended method of installing ESX.
❖ For cases where the mouse, the keyboard, or the video adapter does not func-
tion properly, a text-based installation interface is also available. Some system
administrators prefer the text-based installation to the GUI-based one.
Follow these high-level steps to install ESX from CD using either the graphical or the
text-based installation option:
1. Insert the ESX boot CD into the CD-ROM drive and power on the
system; ensure the system is configured to boot from CD. The system
commences its boot process and the installation mode selection page is
displayed.
2. Press Enter to start the graphical installer or type “esx text” and press Enter
to start the text-based installation.
3. Accept the VMware license agreement.
4. Select the appropriate mouse and keyboard.
5. Add any necessary custom drivers.


6. Select “Yes” to load the system drivers.
7. Enter a serial number, or leave blank to centrally administer serial numbers
in vCenter Server. The system will run in evaluation mode for 60 days.
8. Select the correct network adapter and specify a VLAN ID if required.
9. Configure the network settings on the next screen. Configure the ESX
host IP address by entering a static IP address or a DHCP-based IP
address. (VMware recommends using a static IP address.) Specify a host
name and an optional VLAN ID as well. It is recommended to test the
chosen settings.
10. Select the Advanced setup type. This enables customization of storage
partitions.
11. ESX is installed on the first drive that is seen by the system by default,
either a local SCSI drive or a Fibre Channel/iSCSI LUN. If selecting the
latter, you perform a boot-from-SAN installation. This drive (or LUN) is
initialized and partitioned during the installation process. Keep in mind
that during initialization the contents of the disk will be erased.
12. Each VMware ESX Server will store its Service Console VMDK on a
VMFS volume. This VMFS volume will by default be named “Storage1”.
We recommend preceding Storage1 with the host name of the server. This
makes the volume unique within vCenter Server.
13. The default partitioning scheme is basic and only contains partitions for
swap, /var/log, and root. Our preferred disk-partitioning scheme for ESX
is illustrated in Table 1.
/dev/sda (Primary Partition)
Mount Point Partition Type Size Description
/boot ext3 1100 MB Holds the VMkernel binaries
N/A vmkcore 150 MB Pre-configured
N/A VMFS Remain- For local VMFS volume
ing
esxconsole.vmdk
The root partition for the Service
/ ext3 5120 MB
Console
/tmp ext3 2048 Used to store temporary files
N/A swap 1600 MB 2x maximum Service Console memo-
ry (800 MB)
/home ext3 2048 MB Used for storage by individual users
Used for log files from 3rd-party
/opt ext3 2048 MB
agents
/var/ ext3 5120 MB Separate to avoid overfilling root with
log and core files
Table 1: ESX Server 4.x partition table
14. Partitioning changed with VMware vSphere. The Service Console
is isolated in a VMDK which is hosted on a VMFS volume. During
installation only three physical partitions will be created: physical /boot,


vmkcore, and a VMFS volume. This VMFS volume will hold the Service
Console VMDK which will, in turn, hold all remaining virtual partitions.
15. Specify on the next screen that the ESX is to boot from the drive.
16. Provide boot options for the installer:
a. General kernel parameters—Use this option to add default boot options
to the boot command. These options are passed to the ESX kernel every
time it boots.
b. Force LBA32—This option is used to exceed the 1024 cylinder limit
for the /boot partition. This option is only needed for legacy hardware.
17. Select the time zone.
18. Manually set the time or configure NTP servers.
19. Choose a secure root password and enter it twice in the fields provided.
(Secure passwords usually consist of at least eight characters of which a
combination of letters, numbers, and symbols is preferred.)
20. Confirm your installation configuration and click Next. The installer
starts the installation of the ESX software. Progress bars appear to show
the status of the installation, and a dialog box provides a notification of
completion.
21. Click “Finish” to exit the installation.

Kickstart-Scripted Installation
ESX software can also be installed by leveraging Red Hat’s Kickstart installation method.
Kickstart allows for the quick installation of ESX without having to go through all the
installation screens. The workflow for a Kickstart-based install of ESX is straightforward.
1. Boot the server on which ESX is to be installed, either over the network
using PXE (Pre-boot Execution Environment), a boot floppy, or a boot
CD.
2. The boot program reads a specially created Kickstart configuration file
and determines the preferred installation method (FTP, NFS, or HTTP).
The Kickstart file automatically answers prompts for information such as
network parameters and partition sizing.
3. The installation procedure continues unattended until ESX is completely
installed.
We generally recommend using a solution that is easy to manage and repeatable. Ex-
amples of this would be the Ultimate Deployment Appliances and the ESX Deployment
Appliance. EDA can be found at http://www.vmware.com/appliances/directory/89313/,
and UDA can be found at http://www.ultimatedeployment.org/. Both are appliances
that can be deployed within a matter of minutes. Examples of scripted installations can
be found at http://communities.vmware.com/thread/90424?tstart=120.
See Host Profiles for a vSphere advanced tool for provisioning ESX hosts and maintain-
ing host consistency within the infrastructure.

VMware vCenter Installation
A VMware vCenter Server configured with the hardware minimums can support 15
concurrent clients, 50 ESX hosts, and 250 powered-on virtual machines. A single 64-bit


Windows OS-based quad-processor VMware vCenter Server with 8GB RAM can scale
to 30 concurrent client connections, 300 ESX hosts, and 3000 virtual machines. With
Linked Mode enabled, each environment can scale up to 10,000 powered-on virtual ma-
chines and 15,000 registered virtual machines. The current limit is 10 vCenter Servers.
VMware vCenter Server version 4 has the following prerequisites for successful installa-
tion and usage (see the documentation for the version you plan to install).

Hardware Requirements
❖ Processor: 2GHz or faster Intel or AMD x86 processor. Processor requirements
can be larger if your database is run on the same hardware.
❖ Memory: 3GB RAM minimum. RAM requirements can be larger if your data-
base is run on the same hardware.
❖ Disk storage: 2GB minimum, 3GB recommended. You must have 601MB free
on the destination drive for installation of the program, and you must have
375MB free on the drive containing your temp directory. Additionally, you
will need 1.13GB free on the c: drive for Microsoft .NET 3.0 SP1, Microsoft
ADAM, and Microsoft Visual C++ 2005 Redistributable.
❖ Database storage requirements: The demonstration database, using Microsoft
SQL Server Express 2005, requires up to 2GB free disk space to decompress
the installation archive. However, approximately 1.5GB of these files are delet-
ed after the installation is complete. Note that MS SQL and Oracle are recom-
mended for production use. SQL Express is supported for production use but
should only be used for small environments (up to five hosts).
❖ Networking: Gigabit recommended.

Software Requirements
The VMware vCenter Server is supported as a service on these operating systems:
❖ Windows XP Pro SP2
❖ Windows Server 2003 (all releases except 64-bit), Windows Server 2003 R2
❖ Windows 2008

Database Configuration
VMware vCenter Server supports the following database formats:
❖ Microsoft SQL Server 2005 (SP1, SP2, and SP3 only) and Microsoft SQL
Server 2008
❖ Oracle 9iR2, 10gR1 (versions 10.1.0.3 and higher only), 10gR2, 11g
❖ Microsoft SQL Server 2005 Express
❖ With Update 1, vSphere includes support for IBM DB2

Each database requires some configuration adjustments in addition to the basic installa-
tion. Please see the ESX and vCenter Server installation guide for further details.
The VMware vCenter Server database can coexist with the VMware vCenter Server on
the same physical or virtual machine provided there are sufficient compute resources to


handle the load. However, in an enterprise production setting, it is advisable to host the
VMware vCenter Server database on an enterprise database server that is monitored,
maintained, and regularly backed up. This ensures prompt recovery in case of data loss
or database failure.

License Server Configuration
The license server as it existed in VI3 has been deprecated and is only needed for back-
wards compatibility in mixed environments. VMware vSphere licenses are maintained
within vCenter and locally cached on each host. This avoids the unnecessary downtime
that occurred in the VI3 timeframe when the license server was unavailable for longer
than 14 days. An evaluation license is bundled with vSphere which provides 60 days of
use.

Installation
The VMware vCenter Server is a Windows executable and is installed as a Windows
service. The server can be installed on a physical or virtual machine. While installing the
VMware vCenter Server, you may opt to install the VMware vCenter Server database on
the same physical or virtual machine as the VMware vCenter Server.
The user installing VMware vCenter Server requires administrative privileges on the ma-
chine. For detailed installation instructions, see the ESX and vCenter installation guide.
After the VMware vCenter Server is installed, the features specific to VMware vCenter
Server can only be enabled by adding the appropriate licenses. Unlike ESX, which can
be enabled by a host license file, the VMware vCenter Server features can only be used
with licenses served by vCenter itself.
A default installation of VMware vCenter includes VMware vCenter Orchestrator.
vCenter Update Manager, vCenter Guided Consolidation, vCenter Converter Enter-
prise, and the vCenter client can be installed from the autorun.exe application available
on the installation CD.
The look and feel of VMware vCenter Server through the vSphere Client has been vastly
improved by introducing a new home screen which simplifies administration.

Figure 24: vSphere Client view of inventory, administration, and management

7. Managing VMware vSphere

VMware vCenter Server
VMware vCenter Server is the VMware vSphere management software. It provides a
single point of control for all VMware vSphere resources including ESX hosts, virtual
machines, virtual networking, and virtual storage.
VMware vCenter Server is accessed through the VMware vSphere Client. The interface
provides administrators with a view of the entire virtualized datacenter and the means to
manage its resources.

Figure 25: vSphere Client view of vCenter Server operations
ESX provides a robust virtualization layer, but it is VMware vCenter Server that provides
the features required for a complete infrastructure management system. VMware vCen-
ter Server enables virtual machine mobility, centralized management, centralized moni-
toring and alerting, advanced quality of service, and high availability functionality which
moves ESX from a tactical solution to an enterprise strategic infrastructure platform.
The following sections describe the VMware vCenter Server features in greater detail.

VMotion
A virtual machine can be migrated between two ESX hosts in one of two ways. The first
option is to power off the virtual machine, move it to another host, then power it back
on. This is called a cold migration. The second option is to move a virtual machine be-
tween hosts using VMotion.
VMotion offers huge benefits to datacenter administrators, who can now move run-
ning virtual servers to another host when it is necessary to perform maintenance on an
ESX host. With the help of this technology, each virtual machine can be considered as a
workload, and the workloads can be load-balanced across physical hosts. DRS technolo-
gy does this automatically, utilizing VMotion to load-balance a VMware vSphere cluster.

54 / Managing VMware vSphere

One of the prerequisites for VMotion is that the virtual disk and the other files that en-
capsulate the virtual machine exist on shared storage such as SAN, iSCSI, or NAS. Also,
these files must be visible to both physical hosts between which the virtual machine is
being migrated.
Successful migration of virtual machines between hosts using VMotion requires identical
CPU instruction sets on both hosts, accomplished with either utilizing hosts with the
same family of CPUs or using Enhanced VMotion Compatibility (EVC) on a cluster of
hosts with technologies enabling VMotion compatibility with older servers. More infor-
mation on this and other VMotion requirements is available on the VMware Web site.
VMware vCenter enables VMotion migration or cold migration by dragging the virtual
machine and dropping it on the destination host. A wizard opens to provide guidance
through the migration process. Another option is to use the context menu by right-
clicking on the virtual machine you want to move and selecting “migrate.” This also
opens the migration wizard for guidance through the process.

Virtual Machine Provisioning
Physical servers can be virtualized using a variety of P2V tools and techniques. Admin-
istrators can also opt to build a virtual server from scratch by creating a new virtual
machine, installing the operating system, installing all required services and applications,
and patching the operating system and applications.
After creating a virtual machine by virtualizing a physical server or by building a new
virtual server in the virtual environment, VMware vCenter Server enables marking a vir-
tual machine as a template, which is a virtual machine that cannot be turned on. Tem-
plate creation takes place after installation of the required applications is complete and
is accomplished by creating a VM instance, right-clicking on the virtual machine in the
VMware vCenter Server, and selecting “Convert to Template.” An unlimited number
of virtual machines can be created from this template by right-clicking on the template
and choosing “Clone to Virtual Machine.” This method of creating a virtual machine
is called provisioning from a template. The guest operating system (computer name, do-
main, and IP address) can be customized when provisioning from a template by manu-
ally entering the customization information or by using Sysprep. For more information,
see the Basic System Administration guide.
Provisioning from a template is an invaluable VMware vCenter feature and significantly
reduces the time required to create a new server. Administrators can create different
templates for different purposes. For example, a Windows 2003 Server template might
be created for the finance department, a Red Hat Linux template for the engineering
department, and a Windows 2008 Server template for the marketing department. The
administrator can quickly provision a correctly configured virtual server on demand.
The ease and flexibility of template creation bring the risk of virtual machine sprawl,
where virtual machines are provisioned so rapidly that documenting and managing the
virtual machine lifecycle becomes a challenge. Many VMware vSphere 4 customers are
implementing change management processes in the virtual environment. It is helpful to
have virtual machine naming conventions that help identify each machine’s purpose and
users. VMware Lifecycle Manager can significantly ease the burden with both workflow

Managing VMware vSphere / 55

approval processes and by assigning limited lifecycles to certain VMs or categories of
VMs.

Virtual Machine and ESX Monitoring
VMware vCenter Server provides a way to monitor the performance statistics of the
virtual machines and ESX hosts. The CPU, memory, network, and disk parameters
displayed on the Performance tab of the vSphere Client can be customized and graphed
from short-term to yearly time frames.
Performance monitoring from within the virtual machine yields incorrect information,
because each virtual machine sees only its share of the resources as dictated by the ESX
kernel (hypervisor). Consequently, any performance monitoring should be done through
ESX. Viewing the performance at a VMware vCenter Server level enables viewing of
metrics for all of the ESX hosts (and their virtual machines) in the environment. The
level of detail and estimated dataset size for performance monitoring can be set by select-
ing the Statistics setting in the VMware vCenter Server Configuration window accessible
from the Administration menu in the VMware vSphere Client.
Alarms can be set to trigger when certain metrics exceed a set threshold. For example,
an alarm can be defined to be triggered when CPU utilization on a host reaches 90% or
when the health of the host hardware has changed. New alarms can be created by select-
ing the object (ESX host or virtual machine), clicking on the Alarms tab, right-clicking
on the Alarms panel, and selecting New Alarm.
The action that VMware vCenter Server should take in case an alarm is triggered can
also be defined. The available actions for an ESX host are to send a notification email
message, send a notification trap, enter maintenance mode, exit maintenance mode,
enter standby, exit standby, reboot, or run a script. When creating an Alarm at a virtual
machine level, the virtual machine can be suspended, powered off or reset. Using the
Alarms feature avoids the need for third-party tools that run in the virtual machine or in
the service console of ESX, consuming CPU cycles.

Intelligent Clusters
VMware vCenter Server 4 comes with the ability to create a cluster, a logical group of
servers that can be managed together and work in tandem for load balancing or disaster
recovery.

Figure 26: Stand-alone host versus a vCenter ESX cluster
As servers are added to a cluster, the available resources in the cluster accumulate. Virtual
machines created on an ESX host are displayed in the cluster, not in the individual host.
This is an important distinction. Although a virtual machine in a cluster can exist on
only one host at a time, the virtual machine can execute on any ESX host in that cluster.
VMware vCenter Server can move the virtual machine (or workload) using VMotion


initiated by DRS, or VMware HA, to a different host for the purpose of load balancing
or recovering from a physical host failure.
A cluster in VMware vCenter Server can be configured to balance its load among hosts
by moving its virtual machines as necessary. VMware DRS uses VMotion to distribute
the workload in the cluster. DRS can be configured to automatically use VMotion to
migrate its virtual machines between hosts when the need to load balance occurs or to
prompt the administrator with a recommendation to move the virtual machine over to
the new host. The former configuration is called fully automated mode and the latter is
called manual mode.
VMware High Availability is another feature of VMware vCenter Server clusters. A clus-
ter with VMware HA enabled automatically reacts to a host isolation or host failure by
powering on a virtual machine on an alternate good host. This ensures that, in case of
an unplanned host failure, the virtual machines are brought back to life without manual
intervention.
VMware HA does not use VMotion to migrate virtual machines to a good host. The
virtual machines are restarted on the alternate host, so the virtual machine users will
notice a temporary service interruption. VMware HA is a reactive feature triggered by
ESX isolation or failure, while DRS is a proactive approach to keep the load on the hosts
balanced at all times. VMware HA, once configured, does not require VMware vCenter
Server to operate. The ESX hosts communicate directly with the other ESX hosts in the
cluster to determine whether a particular ESX host is offline. VMware vCenter Server
can be run within a virtual machine and the VMware HA technology will still function
and restart the VMware vCenter Server along with the other virtual machines.
VMware HA responds to host isolations and host failures. VMware vSphere offers two
high availability features which enable per-virtual-machine-level high availability. VM
Monitoring is an integral part of VMware HA. VM Monitoring restarts individual virtu-
al machines when VMware Tools heartbeats are not received within the given threshold.
Like VMware HA, this is a form of stateless clustering, meaning that the virtual machine
will be restarted instead of a seamless failover.
VMware FT provides a stateful solution; that is, when VMware FT is enabled on a spe-
cific machine and the host that is running the primary virtual machine fails, the second-
ary virtual machine seamlessly takes over. VMware FT provides continuous availability
by keeping a shadow copy of a virtual machine in lockstep with the primary virtual
machine. Like VMware HA and VM Monitoring, VMware FT also uses heartbeats to
monitor the status of a virtual machine and instantly replaces the primary with the sec-
ondary when heartbeats are lost.
A VMware vCenter Server cluster can be configured for both VMware DRS (including
DPM) and VMware HA (including VM Monitoring) and can recover from unplanned
host isolation and host failures. Within a VMware vCenter Server cluster, VMware FT
can be enabled for virtual machines which meet all prerequisites and which need con-
tinuous availability.

Managing Resource Pools
VMware vCenter Server enables carving of resources for a set of virtual machines. For
example, 25% of the CPU resources and 30% of the memory resources might be as-


signed in a cluster to a set of staging virtual machines. This reservation is called a resource
pool. An additional resource pool can be created on the same cluster with 70% of the
CPU resources and 60% of the memory resources for production virtual machines.

Figure 27: VMs assigned to resource pools representing shares of the physical re-
sources
This configuration ensures that the staging virtual machines never use more than 25%
of the CPU resources on the cluster. A rogue staging virtual machine that is consuming
excess CPU cycles cannot affect the production virtual machines.
Resource pools also help organizations that use an IT chargeback model to limit VM
resource use by a particular department to only those resources allocated to that depart-
ment.
The combined effect of resource pools, DRS, DPM, and VMware HA is a radical shift
in how administrators think about their datacenter. The clusters in a datacenter can
now be considered a pool of compute resources. Administrators can carve out portions
of these resources for a set of virtual machines depending on their service level require-
ments.

VMware Capacity Planner
VMware Capacity Planner is described in the section “Identifying Virtualization Can-
didates” in Chapter 5. This is an add-on module for VMware vCenter Server 4 which
collects comprehensive resource utilization data in heterogeneous IT environments. It
compares the information gathered with industry-standard reference data to provide
both analysis and decision support modeling. Scenario modeling provides guidance in
the design of a virtual infrastructure.

VMware Update Manager
VMware Update Manager is part of the VMware vSphere 4 product suite. It supports a
streamlined, automated update mechanism for ESX hosts and virtual machines. See the
VMware Update Manager section later in this chapter.

VMware Guided Consolidation and VMware Converter
VMware Converter is described in the section VMware Guided Consolidation and VM-
ware Converter in Chapter 8. This is an add-on module for VMware vCenter Server 4.


Distributed Power Management
VMware Distributed Power Management (DPM) is a new technology as of ESX 3.5
which supports dynamic utilization of ESX hosts. The initial release was considered
experimental; the current version is fully supported. Based on virtual machine resource
requirements using DRS, ESX hosts are powered on or off to support virtual machine
requirements and maximize power savings during periods of lower resource require-
ments. This enables a minimal number of ESX hosts to be running to support the re-
source requirements.

Infrastructure Management with VMware vCenter
Effective management tools and processes are essential components of a VMware
vSphere environment. Hardware selection and configuration determine, to a large ex-
tent, the performance and reliability of the infrastructure.
Ongoing change is common in virtualized environments. One of the strengths of a VM-
ware vSphere environment is the ability to create, change, and reallocate virtual resources
in real time independent of the management of the underlying physical hardware. Some
of these changes may happen automatically in order to maintain the environment. This
dynamic environment and ease of change demand a comprehensive tool set capable of
providing administrators with intuitive interfaces that enhance and integrate with stan-
dard support processes and tools.
VMware vCenter Server provides administrators with a single management system for
controlling and monitoring the virtual servers, networks, storage, and physical host serv-
ers as a unified infrastructure. The ability to consolidate and standardize infrastructure
management into unified tool and set processes is one of the driving features for infra-
structure agility. Integrating all infrastructure management tasks through the use of roles
and views helps drive the long-term ROI and change in an organization’s service support
and management. A poorly managed virtual infrastructure can erode the ROI and un-
dermine the stability of the infrastructure.
The following are some of the VMware vCenter interfaces that are routinely used for
VMware vSphere management tasks.

Datacenters and Clusters
The Hosts and Clusters page is available from the home page in VMware vSphere Cli-
ent. This interface is where the physical and logical organization of the VMware virtual
infrastructure occurs. Datacenters, folders, and clusters are containers for organizing
VMware vSphere components. The datacenter object is used to represent a physical
datacenter or location. Each datacenter created is the root of a hierarchy of folders or
containers organizing the hosts, clusters, virtual machines, networks, and datastores
within it.
The datacenter is a logical isolation boundary for a set of hosts, networks, datastores,
and datacenters online with VMotion or offline with a cold migration. Templates are
specific to a datacenter and cannot be shared, and all object names within a datacenter
must be unique.


VMware vSphere clusters are created within a datacenter object to aggregate the re-
sources and management of a number of ESX hosts. Clusters are the main management
containers in VMware vCenter. Virtual machines and resources are distributed and man-
aged within clusters. DRS (including DPM) and VMware HA services are managed and
activated at the cluster level. Hosts are placed in a cluster, where they are managed as a
logical unit. Host resources are aggregated within the cluster.

Roles and Permissions
Centralized security and permissions for the VMware vSphere infrastructure are stored
and managed in the VMware vSphere Client through the Permissions tab. User accounts
and permissions for local host access can be assigned to the service console or the VM-
ware vSphere Client when connected directly to an ESX host.
Roles are used to define privileges—that is, a set of individual rights to read properties
or to perform actions within VMware vCenter. Users and groups from a Windows Ac-
tive Directory are placed in roles. Permissions are created by assigning roles to VMware
vCenter Server objects. Members of the VMware vCenter Server local Windows admin-
istrators group have the administrator role in the VMware vCenter Server.
The propagation of permissions through hierarchies is optional and is set for each per-
mission rule. Permissions set on a VMware vCenter Server object override permissions
set on the parent object.
VMware vCenter Server and ESX offer the following predefined roles:
❖ No access
❖ Read-only user
❖ Administrator
❖ Virtual machine user
❖ Virtual machine power user
❖ Resource pool administrator
❖ VMware Consolidated Backup user
❖ Datastore consumer
❖ Network consumer
Roles can be created, edited, removed, renamed, or cloned using the VMware vSphere
Client through the Roles page on via the Home view by right-clicking in the Roles panel
and choosing Add Role.
The combination of VMware vCenter roles and the nested hierarchies of VMware
vSphere and DRS makes it simple to delegate management of virtual resources to groups
of users. This is an effective way to provide localized resource and service-level control to
groups of users while reaping the benefits of infrastructure consolidation.
The most effective way to manage permissions through VMware vCenter is to create
resource pools for groups of virtual machines for management and resource sharing,
and Active Directory groups for each resource pool. By assigning each Active Directory
group permissions to the appropriate resource pools, VMware vCenter permissions can
be managed through standard Active Directory security practices.


Events
Any event of interest in a VMware vCenter Server or an ESX host triggers an event mes-
sage. Event messages are stored in the VMware vCenter Server database and are viewable
in the VMware vSphere Client. The Events page, accessed via the Home page, on the
displays all of the events in the VMware vSphere infrastructure, and the Events tab in
the inventory for any object in the hierarchy enables you to see the events for that par-
ticular virtual machine, host, or datacenter.
Events are displayed in scrolling chronological order and are color-coded for Informa-
tion, Error, or Warning messages. Event details can be seen by selecting the specific event
and viewing the Event Details window.
The VMware vSphere Client enables greater visibility and event management by allow-
ing sorting of event columns or searching for events. These features are necessary when
reporting on events in large, consolidated environments. Events are searched for using
the event-filtering feature in the VMware vSphere Client, allowing individual events to
be identified in a contextual view.

Figure 28: vSphere Client view of events

Alarms
VMware vCenter Server and ESX provide a comprehensive set of tools for monitor-
ing and triggering alarms in response to conditions or events within the virtual infra-
structure. Default alarms are included in VMware vCenter, but alarms are not directly
configurable through the VMware vSphere Client connected to ESX. In addition to the
default alarm settings in the VMware vCenter Server, a rich interface is available for cre-
ating additional alarms that can alert on some or all inventory objects by being applied
to objects or containers in the VMware vSphere hierarchy. VMware vSphere introduced
many new events, triggers, and actions which are invaluable for system administrators.
VMware vSphere makes it possible to initiate maintenance mode when the hardware
health status of a host degrades. When a host is placed in maintenance mode, all virtual


machines will automatically be VMotioned to all other available hosts in the cluster and
avoid possible downtime because of hardware outage.

VMware vCenter Server Alarms Tab
VMware vCenter Server provides two main views for setting and monitoring alarms,
the Alarms view and the Definitions view. These views are available from the VMware
vSphere Client connected to a VMware vCenter Server, Home: Hosts & Clusters > Host
> Alarms tab: Alarms or Definitions buttons.
The Alarms view displays any active alarms that are triggered. The Definitions view is
where existing alarms are listed and where alarms can be edited or added.
Alarms trigger actions when the event defined in the alarm occurs on a host or virtual
machine. Alarms are applied to datacenters, folders, clusters, resource pools, hosts, or
virtual machines to display or notify regarding the status of one or more of these objects.
Alarms defined within the hierarchy are inherited by all of the object’s child items. This
cascading of alarms within the hierarchy cannot be prevented.
Alarms can be defined or edited by a user with the appropriate permissions on datacen-
ters, hosts, and any virtual machines within the scope of the alarm definition, that is, the
object the alarm is applied to and the flows to its child objects.
Alarm triggers define conditions that, when met, cause an event to be activated. Alarm
triggers can be defined as percentages above or below host or virtual machine memory or
CPU utilization, or by a virtual machine’s heartbeat. Alarms can also be triggered based
on the state of a host or virtual machine. State alarms define a state as “Is” or “Is Not” a
particular condition.
Examples of virtual machine and host condition states are:
❖ Virtual machine powered on
❖ Virtual machine powered off
❖ Virtual machine suspended
❖ Host powered on
❖ Host powered off
❖ Host suspended
New condition states can be defined for datastores:
❖ Connected to all hosts
❖ Disconnected from all hosts
Alarm notifications are actions taken when an alarm is triggered. Notification actions
available in VMware vCenter Server include the following:
❖ Send an email notification message.
❖ Send an SNMP notification trap.
❖ Run a command.
❖ Enter maintenance mode.
❖ Exit maintenance mode.
❖ Enter standby.


❖ Exit standby.
❖ Reboot host.
❖ Suspend a virtual machine.
❖ Power off a virtual machine.
❖ Reset a virtual machine.

VMware vCenter Server alarms and notifications provide a rich set of tools for notifying
and reacting to alarm events within the infrastructure, and they allow seamless integra-
tion of a VMware vSphere infrastructure with IT Service Management tools and prac-
tices.

Virtual Machine Deployment
Virtual machine deployment and management are among the most common manage-
ment tasks for VMware vSphere administrators. The VMware vSphere Client provides
interfaces for virtual machine management that are integrated with the alerting, moni-
toring, and organizing views.
Virtual machines can be created by manually installing an operating system, by using
a preconfigured template, by importing an appliance, by cloning an existing virtual
machine, or by importing a physical server or a virtual server from another hosting plat-
form.
A simple wizard-based interface is used to create a new virtual machine manually from
a template or by cloning an existing virtual machine. Virtual machine deployment is an
area where updating organizational processes to accommodate the VMware vSphere in-
frastructure is important to maintaining availability and provides significant operational
advantages.
Resource groups and distributed permissions give customers or groups within the orga-
nization the ability to create and manage virtual machines within their resource group.
Moving these tasks closer to the owners by making the Create New Virtual Machine
wizard available (through the VMware vSphere Client or, preferably, through an online
service catalog or portal) reduces overhead on the infrastructure management team and
increases the speed and accuracy of server request fulfillment.
Reducing server deployment lead time and distributing access to the infrastructure re-
quire enhanced capacity and change management functions. The VMware vSphere Cli-
ent provides simple centralized monitoring of the VMware vSphere infrastructure capac-
ity and the ability to non-disruptively add and distribute additional capacity throughout
the infrastructure. Organizations with large virtual infrastructures need to merge this
manageability and agility with their service and support processes for maximum benefit.
The information needed to build a virtual server is similar to the information gathered in
project planning for physical server acquisition. New virtual servers can be created from
the Summary tab or from the context menu of host servers, clusters, or resource pools in
the VMware vSphere Client. The Create New Virtual Machine wizard performs the fol-
lowing information-collecting activities for virtual server creation:
❖ Selects a virtual machine name. This is not the domain name of the server, but
they can be the same. The name must be unique within the datacenter context;
it is used to name the virtual machine’s files.


❖ Selects a location for the virtual machine in a folder, resource pool, or datacen-
ter.
❖ Selects a datastore for the virtual machine files.
❖ Identifies the guest operating system to be installed on the virtual machine.
❖ Sets the virtual disk size. A virtual disk can be from 1MB to almost 2TB in size
and either thin or thick. Additional virtual disks can be added later. The size
of existing virtual disks can be changed through the VMware vSphere Client.
When cloning an existing virtual machine or creating one from a template,
only the information that differentiates the new virtual machine from the tem-
plate or existing server is required.
Creating a virtual machine using the custom option includes several more detailed op-
tions: amount of vCPUs, amount of memory, amount and type of network adapters,
type of SCSI adapter, LSI Logic Parallel or SAS, BusLogic Parallel, or VMware Para-
virtual. Select the type of virtual disk: new, existing, or a mapped LUN from a storage
network. When creating a new disk, virtual disk mode can be specified. The disk modes
allow for persistent disks, which commit changes to disk immediately, and non-persistent
disks, on which changes are discarded when the server is powered off or reverted to a
snapshot.
To map a raw SAN LUN to a virtual machine:
❖ Select a target LUN visible from the SAN.
❖ Choose a datastore.
❖ Choose physical or virtual disk compatibility mode.
❖ Physical compatibility mode gives the guest operating system direct access
to the disk hardware. Disks in physical compatibility mode have the restric-
tions of not being able to be cloned, migrated, or used for templates. This
is used for application clustering technologies (MSCS and VCS) and for
utilizing vendor-based storage management tools.
❖ Virtual compatibility mode maps a SAN LUN to the guest OS, but re-
tains the functionality of a virtual disk.
After completing the Create New Virtual Machine wizard, you can choose to edit the
virtual machine setting before completion. This enables adding or removing specific de-
vices or setting reservations and limits.
Guest operating systems can be installed into a new virtual machine by mapping a CD,
DVD, or disk image to the virtual machine and booting to that device. After it is booted
to the installation disk, operating system installation can continue as usual.
The boot order for a virtual machine can be altered by pressing F2 during initial post-
test of the machine or by checking the “Enter BIOS on next boot” box in the virtual
machine options. This might be necessary to enable boot from CD or DVD functional-
ity.
After installation of the guest operating system, the VMware Tools service should be
installed to allow full manageability of the virtual machine. VMware Tools includes driv-
ers for the virtualized hardware and scripts to enable shutting down, suspending, and
restarting of the virtual machine from the VMware vSphere Client.


Migration of Virtual Machines to Alternate Platforms
It is possible to migrate virtual machines from one virtualization platform to another.
There are a number of ways to do this, depending on the source and target virtualization
platforms.
VMware Converter handles migrations between ESX hosts, VMware Server, and VM-
ware Workstation. VMware Converter can also import from other Intel/AMD-based
virtualization platforms such as Microsoft Virtual Server and Virtual PC virtual ma-
chines from Symantec Backup Exec System Recovery, Symantec LiveState, StorageCraft
ShadowProtect, Norton Ghost images, Acronis True Image, and VMware Consolidated
Backup full backup images.

Hot Migrations (VMotion)
Virtual machines can be migrated between hosts within a datacenter for performance,
resource balancing, or maintenance reasons. VMotion migrations, also known as hot mi-
grations, allow a virtual machine to be moved from one virtualization platform to another
with no downtime. The virtual machine migration wizard can be started from the con-
text menu or the Summary page of a virtual machine. VMotion migrations can also be
triggered by automated DRS operations to balance resource usage in resource pools.
VMware vCenter Server validates the virtual machine and its target destination host and
resource pool to ensure a successful migration. Hot migrations of virtual machines can
occur only when certain conditions are met, such as when the start host and destination
host have compatible CPUs or Enhanced VMotion Compatibility has been enabled.
Both hosts must have access to the same datastores and adequate resources to perform
the migration without violating any resource pool admission requirements.
Migration validation warnings are issued if the virtual networks on the host and destina-
tion sites do not match. Non-matching networks during a VMotion migration cause the
guest OS to be disconnected from the network when the migration is complete, requir-
ing a new network and NIC to be assigned and connected at the destination side of the
migration.
VMotion migrations fail validation if removable devices such as CDs, DVDs, or floppy
drives are mounted and connected to host or client hardware. Disconnecting or unload-
ing these devices eliminates the problem and allows VMotion migration to occur if all
other requirements are met.
The speed and success of VMotion migrations depend upon the active load of the host
and destination servers and the activity of the virtual machine to be migrated.

Figure 29: VMotion


Storage VMotion
Storage VMotion is based on the hot migration technology of VMotion, but also enables
moving a virtual machine’s virtual disk to a different datastore. This is available with
ESX 3 and newer versions with the appropriate version of VMware vCenter Server. An
example is migrating a running virtual machine from one ESX host to another while
also migrating the location of the virtual machine’s virtual disk to another storage loca-
tion. Another example would be migrating the virtual disk for the virtual machine from
one LUN to another without moving the virtual machine to another ESX platform.
Storage VMotion is supported for virtual machines stored on Fibre Channel and iSCSI
SAN shared storage or NFS shares. It performs proactive, non-disruptive storage migra-
tions to simplify array storage maintenance without virtual machine downtime, and it
enables changing from thin-provisioned to thick-provisioned disks or vice versa. Storage
VMotion can be used to eliminate storage I/O bottlenecks without impacting virtual
machines, just as the standard VMotion eliminates CPU bottlenecks. Storage VMotion
uses existing VMware technologies such as changed block tracking, introduced with
VMware vSphere. The flow of a Storage VMotion migration is as follows:
1. The virtual machine’s home directory (containing virtual machine
configuration, swap, and log files) is copied to the new storage location.
2. Optionally: The virtual machine is migrated, using VMotion, to the target
ESX host.
3. Change block tracking is enabled on the source VMDKs.
4. The parent virtual disk is copied from the old storage device to the new
storage device via the Data Mover, which enhances performance and
decreases overhead on both source and destination storage devices.
5. All remaining changed blocks are copied from the old storage device to the
new storage device.
6. The source VM is suspended and the destination VM is resumed when
the state (checkpoint data) is copied to start running on the new home
and VMDKs.
7. The source VM home and VMDKs are deleted.
The entire process takes several minutes, depending on the size of the VMDKs of the
virtual machine being migrated.

Figure 30: Storage VMotion


Cold Migrations
Cold migrations perform the same function as Storage VMotion for virtual machines
that are powered off. Cold migrations allow the movement of a virtual machine that is
powered off from one physical ESX host to an alternate physical ESX host. Cold migra-
tions, unlike VMotion, allow movement between ESX hosts utilizing different vendor
types and CPU families without masking specific CPU characteristics.

VMware Update Manager
VMware Update Manager provides an automated patch management system for VM-
ware vSphere. The software manages the tracking and patching of ESX. It also handles
select Windows and Linux virtual machines.
The VMware Update Manager eliminates the risk of manually tracking and patching
ESX and virtual machines by the following strategies:
❖ Scanning the state of the physical VMware ESX hosts.
❖ Scanning the state of select guest operating systems running within the virtual
machines.
❖ Scanning the state of select applications running within the virtual machines.
❖ Comparing the current state of each with baselines set by the IT team.
❖ Applying updates and patches for compliance with business requirements.
❖ Allowing for a snapshot prior to patching to enable a method to back out of the
patch, if needed.
❖ Allowing patching of offline virtual machines.
❖ Working with DRS to eliminate disruptions during ESX host patching:
❖ Hosts are placed in maintenance mode one at a time.
❖ Virtual machines are migrated to other hosts prior to patching and are mi-
grated back after patching is complete.
Updates to the ESX do not introduce virtual machine downtime.
VMware Update Manager provides notification of host patches and allows update sched-
uling. It also handles all of the virtual machine migration, patching, and rebooting, elim-
inating typical manual steps and downtime during updates in a physical environment.

VMware vCenter Orchestrator
VMware vCenter Orchestrator is an orchestration engine simplifying management with
task automation. VMware vCenter Orchestrator relies on workflows to execute auto-
mated processes. These workflows can contain specific actions or decisions based on
output of a specific task and can be reused at any given time, providing consistency in
operational tasks. VMware vCenter Orchestrator provides a library of example workflows
for common management tasks and individual actions.

VMware vCenter Linked Mode
VMware vCenter Server 4 introduces a new feature called Linked Mode. Linked Mode
provides a single point of authentication for a multi-vCenter Server environment as it


logs in simultaneously to all vCenter Servers. Linked Mode simplifies management of
remote offices or multi-datacenter environment by offering a single pane of glass. Roles,
permissions, and licenses are replicated across the virtual infrastructure by utilizing Mi-
crosoft ADAM. The most powerful function, however, is the search option as there are
no boundaries for exploring your environment.

Figure 31: vCenter Linked Mode

VMware Host Profiles
VMware Host Profiles enables centralized compliance monitoring and reporting for host
configuration. VMware Host Profiles can be compared to VMware Update Manager, as
it provides a baseline for configuration items to which the environment needs to comply.
Likewise, VMware Host Profiles can be used to quickly make host configuration changes
to any environment. A single host can be reconfigured and used to capture a blueprint,
which, in its turn, can be applied to a datacenter, cluster, or a single host. VMware Host
Profiles also assists in maintaining the integrity of the host servers in the face of normal
changes over time.
Host Profiles consists of several sub-profiles. Each sub-profile contains policies which
describe the configuration for a specific profile. Examples of these policies are:
❖ Firewall rule sets
❖ DNS settings or updates
❖ Service console memory
❖ NTP configuration
❖ Virtual switch configuration, including NIC teaming and security policies
❖ Additional users or user groups

8. Migrating Virtualization Candidates
Conducting a virtualization assessment helps ensure virtualization project success. After
candidate identification, the migration time can be estimated based on network band-
width, storage requirements, and application complexities.

VMware Physical-to-Virtual Process
The physical-to-virtual (P2V) process is the method used to convert and migrate a
physical server to a virtual machine. VMware uses P2V migration software to handle
data migration and hardware reconfiguration. The process is similar to migrating an OS-
application pairing from one physical platform to another, but instead of changing the
hardware devices and drivers to another physical set, the migration changes them to a
virtual set.
For native virtualization, the virtual set consists of CPU, RAM, SCSI disks, AMD PC-
Net Adaptive NICs, parallel and serial devices, CD-ROM, and floppy devices.
For hosted virtualization, the virtual set consists of all the hardware that the underlying
host’s OS can see. This provides a greater number of devices, including USB and Wintel
modems. Physical machines that are better suited for a hosted virtualization deployment
include fax systems and RAS servers, although over-IP solutions are typically used in
virtualized environments.
Migration checklist:
1. Migrate the OS disk.
2. Reconfigure the OS with new devices.
3. Install VMware Tools.
4. Check for and remove unused devices.
5. Check for and disable unused services.
6. Clean up legacy log information but not warnings or errors that may
assist in troubleshooting application or OS issues once the migration is
complete.
7. Migrate the data disk(s).
8. Reboot.
9. Check logs.
10. If clean, then done. If not, go to step 4.

VMware Guided Consolidation and VMware Converter
VMware Guided Consolidation is an optional vCenter Server component and is used
for planning physical-to-virtual machine migrations. VMware Guided Consolidation
leverages VMware Capacity Planner and VMware Converter technology and is designed
for small-scale consolidation.

Migrating Virtualization Candidates / 69

VMware Converter can convert and migrate physical machines, as well as third-party
disk image formats, to VMware virtual machines. It also converts and migrates virtual
machines from one VMware platform to another.
Local and remote physical machines can be migrated hot or cold. When performing hot
conversions or migrations, there is no downtime for the original physical machine. Mul-
tiple simultaneous conversions and migrations are supported with VMware Converter.
The VMware Converter supports migration from Microsoft Virtual PC and Microsoft
Virtual Server to VMware virtual machines. It also supports migrating from backup im-
ages such as Symantec Backup Exec LiveState Recovery or Symantec Norton Ghost, as
well as restoring VMware Consolidated Backup images to running virtual machines.
VMware Converter supports the import of OVF-based virtual machine images.
VMware Converter Starter and VMware Converter Enterprise both support hot clon-
ing, local conversions, and remote conversions. The differences between the starter and
enterprise versions of VMware Converter include:
❖ The enterprise version allows for cold cloning with a special bootable CD,
multiple simultaneous conversions, automation and management of large-scale
conversions, and remote conversion to all destinations.
❖ The starter version is free, allows single conversions, and limits remote conver-
sion to VMware Workstation, VMware Server, VMware Player, and VMware
GSX Server.

Figure 32: VMware vCenter Converter

70 / Migrating Virtualization Candidates

Third-Party Migration Tools
Several third-party migration tools can be used for migrations. These include commer-
cial products such as Platespin PowerConvert, Leostream P>V Direct, and Ultimate-
P2V (free P2V migration tool based on BartPE Boot-CD).

Manual Migration
Manual migration from physical to virtual machines is possible. Due to the time in-
volved, most businesses choose to utilize automation tools. The manual process is docu-
mented in white papers.
A manual migration follows these high-level steps:
1. Create a helper virtual machine to assist in the migration.
2. Connect additional blank virtual disks to the helper virtual machine to
hold the migrated physical disks.
3. Using imaging or backup software, migrate the physical disk data to the
virtual disks.
4. Modify the boot information to ensure the correct boot order for starting
the virtual machine.
5. Power off the helper virtual machine.
6. Configure a new virtual machine that uses the new virtual disks.
7. Boot the system.
8. Install and configure VMware Tools.

Considerations for Successful Migrations
What constitutes a successful migration? These are the criteria customers rank highest:
❖ A functional guest OS and associated application(s)
❖ Network connectivity
❖ Sign-off by key stakeholders:
❖ Migration staff
❖ OS administrators
❖ Application owners and/or administrators
❖ End users

Virtual-to-Physical Process
The V2P process enables the migration of a virtual machine to a physical server. Micro-
soft Sysprep 1.1 is used to handle the conversion process for Microsoft guest operating
systems. Other options are possible for non-Microsoft operating systems. The Kudzu
imaging tool works very well for Linux environments.
See the V2P TechNote at http://www.vmware.com/support/v2p/doc/V2P_TechNote.pdf.
The note refers to the Kudzu imaging tool but does not specify the functionality. The
following is a short set of steps for migrating a Linux physical machine:
1. Create a new VM and configure appropriate CPU, memory, storage, and
network.

Migrating Virtualization Candidates / 71

2. Boot the physical machine with a real Linux Live-CD disk (e.g.,
Knoppix).
3. Boot the new VM connected to the Linux Live-CD image.
4. Use “dd” to migrate the physical machine’s disk data to the new VM.
5. Boot the new VM in rescue mode.
6. On the new VM use “kudzu” or “yast” to remove old devices and identify
new ones presented to the VM.
7. Update the new VMs init using “mkinitrd.”
8. Modify “grub” on the new VM.
9. Reboot the VM and do final cleanup.

Virtual-to-Virtual Process
VMware Converter allows for migrating from one virtual technology platform to another.
It provides a one-way migration to a VMware virtualization technology such as ESX,
VMware Server, or VMware Workstation.
VMware Converter allows source virtual machines to run on ESX, VMware Server, VM-
ware Workstation, Microsoft Virtual PC/Server, and Xen.
VMware Converter supports the import of OVF-based virtual machine images.

9. Optimization

The consolidation benefits of virtualization result from significantly boosting resource
optimization from the low levels of a physical environment. Often represented as a
percentage such as %CPU or the I/O bandwidth of a system, utilization is the amount
of used resources. Utilization levels normally can be increased without affecting per-
formance, which is measured in the number of transactions that can execute in a given
amount of time, such as per-second Web page views or I/O operations. An exception,
as shown in the following graph, is a negative performance impact as resource capacity
reaches its maximum. The sharing of resources on an ESX host pushes the resource utili-
zation higher when consolidation ratios increase, making it important to understand the
subsequent performance impact.

Figure 33: Performance impact as resource utilization reaches capacity
Performance monitoring enables resource demand balancing to achieve an optimal, or at
least acceptable, performance level.
Applications requiring precise timing, should not be run in virtual machines. This in-
cludes performance monitoring tools, lab measurement applications, and instrument
controllers. The variations introduced to the vCPU through time slicing of the real CPU
can lead to millisecond delays.

ESX Optimization
An individual virtual machine can impact other virtual machines based upon how it uses
resources. Successful consolidation of multiple operating systems and applications under
one set of hardware requires constant monitoring of resource loads and rebalancing of
virtual machines when required, both within and across ESX hosts.

Monitoring
In a physical machine environment, resources are monitored using tools such as Perf-
mon and Disk Manager under MS Windows, and top and vmstat under UNIX.
In a virtual machine environment, guest OS–based performance monitoring tools may
not represent reality, but can nonetheless be useful for understanding application issues.
VMware vSphere tools such as VMware vCenter Server, esxtop, and resxtop (via the

Optimization / 73

rCLI) represent actual performance. Figure 34 is a screenshot of a VMware vCenter cli-
ent showing monitoring of CPU utilization of virtual machines.

Figure 34: Monitoring virtual machine resource utilization and status
VMware vCenter Server is used to view resource utilization and performance metrics.
VMware vCenter Server contains multiple default resource views that can be displayed
on multiple objects such as total CPU utilization and total disk and network I/O. Cus-
tom performance charts can also be created and are invaluable when troubleshooting
performance-related problems.
VMware vCenter provides detailed metrics in real-time for the past hour. Typically a
much reduced set of metrics is available for data more than an hour old. For more de-
tailed metrics, esxtop, vm-support -S, or third-party tools can be used to collect data.

Resource Pools
Resource pools are hierarchical logical containers created within VMware vSphere DRS
clusters. They segment the total cluster CPU and memory resources and/or delegate ad-
ministrative control over the resources and virtual machines within the pool.

Figure 35: VMs assigned to resource pools representing shares of the physical re-
sources
In a VMware vSphere cluster utilizing VMware HA and DRS, all CPU and memory
resources are treated as a logical whole because virtual machines can be executed on any
of the ESX hosts in the cluster, depending on current circumstances. Applying resource
pools to the infrastructure separates the resources from a cluster of servers. Abstract-

74 / Optimization

ing the resource pools from the physical hardware creates a form of virtualized resource
management areas.
Using resource pools within a VMware vSphere cluster provides significant flexibility
in managing resources and distributed administration of virtual resources. VMware
vSphere administrators can delegate complete control of a resource pool to a resource
manager, allowing functional or departmental groups the ability to manage both the
virtual machines and the resource distribution within their managed pool—in essence,
giving the pool administrators a small VMware Virtual Infrastructure of their own. By
allowing both resource sharing and isolation, this feature can be especially useful when
hosting groups of virtual machines for other IT departments.
By default, each ESX host or resource cluster includes an invisible resource group con-
taining the total CPU and memory resources of that host or of all hosts in the DRS
cluster. A resource pool can contain virtual machines or child resource pools utilizing a
portion of the parent pool’s resources. Nesting resource pools within pools creates a par-
ent, child, and sibling resource pool relationship that provides structure for resource dis-
tribution. Using nesting resource pools to delegate administration enables the localized
management necessary for large, consolidated IT environments.
Each resource pool is assigned a reservation, a limit, and a number of shares for CPU
and memory resources. Each resource pool must also be designated as either having an
expandable reservation or not. Combining these settings provides isolation between re-
source pools while allowing sharing within resource pools.
❖ Reservation—A guaranteed amount of CPU or memory resources set aside for
use by the resource pool. Reserved resources are not available for use by virtual
machines in other resource pools. By default this is set to 0.
❖ Limit—The maximum amount of CPU and memory resources available to the
virtual machines and child pools assigned to a resource pool. The default value
for resource pool limits is unlimited.
❖ Shares—The number of CPU or memory shares allocated to the pool from
the parent pool. Resource sharing within sibling pools is done relative to their
share of the parent pool’s total shares. Numerical share values may be compared
only among siblings. For example, the number of shares of a VM in one Re-
source Pool can be compared to the shares for other VMs in the same pool but
not to VMs in other pools. A pool’s share value is constrained by its reservation
and limit. Share values can be set to Low, Normal, High, or Custom. A custom
selection allows a specific share value to be entered instead of selecting one of
the default relative values.
❖ Expandable reservation—This is set by default for newly created resource
pools. Using expandable reservations allows borrowing of resources from a par-
ent (or ancestor) resource pool to start a new virtual machine when the pool’s
existing reservation is exhausted.
Shares, Reservations, and Limits can be applied at the Resource Pool level, the VM
level, or both. Resources are divided at the Resource Pool level first and then subdivided
among the VMs in that pool. For example, if two Resource Pools are at the same level,
one with 100 VMs and the other with 2 VMs, resources are divided between the pools
and then subdivided among 100 VMs for one pool and among 2 VMs for the other.

Optimization / 75

Placing application systems or functionally similar servers in resource pools simplifies
resource management by avoiding the need to individually manage resource allocation to
each server. Resource pools use an admission control system to ensure that the pool limit
is not violated and the reservation is met for all running virtual machines and child pools
contained within it. When a virtual machine is powered on or a child resource pool is
created, admission control checks for available resources in the pool. Insufficient pool
resources, combined with a reservation type set to Fixed prevents powering on the vir-
tual machine. When the reservation type is set to Expandable and the current pool has
insufficient resources, the parent and ancestor pools are checked for available resources.
If sufficient resources are available in one of the parent pools, the resources are borrowed
and reserved by the child pool and the action is allowed.

Figure 36: Resource pool settings
Using expandable reservations requires coordination between the pool administrators, as
a child pool can borrow and reserve all of the resources in a parent or ancestor group.
Resource pools are easily managed from a series of tabbed interfaces in the VMware
vSphere client, similar to the management of hosts or virtual machines.
The combination of fixed and expandable reservations with limits and shares allows
resource pools to validate and ensure service levels for all virtual machines within a VM-
ware vSphere cluster.

Distributed Resource Scheduling
DRS clusters enable load balancing of virtual machines across multiple hosts by creating
logical pools of resources and distributing virtual machines across hosts as required. This
load balancing can be manual, partially automated, or fully automated.
❖ Manual load balancing only provides recommendations for new virtual ma-
chine placement or migrations.
❖ Partially automated provides automatic placement for new virtual machines
being deployed and makes recommendations on migrating virtual machines for
load balancing.

76 / Optimization

❖ Fully automated provides automated placement of new virtual machines and
automatic migration of virtual machines for load balancing. Five modes are
available for this automation level: conservative (5) to aggressive (1).

Virtual Machine Optimization
VMware best practices for single and multi-threaded application CPU allocation rec-
ommend starting with the lowest number of vCPUs when either deploying new virtual
machines or migrating from physical servers to virtual machines. Multi-threaded ap-
plications can benefit from multiple vCPUs, although they require all vCPUs to be
scheduled to perform an execution at the same time. This can reduce the overall com-
puting resources available. Single-threaded applications require only a single vCPU
scheduled for a 4 vCPU VM. The scheduler has been vastly improved with ESX 3.x and
again enhanced with ESX 4.x. However, unnecessary vCPUs still waste physical CPU
resources. Detailed information can be found here: http://www.vmware.com/files/pdf/
perf-vsphere-cpu_scheduler.pdf.

Application and Operating System Tuning
Although a physical server has full access to all resources, it frequently is over-provi-
sioned in order to account for future capacity requirements. An application-OS pair
needing one gigabyte of memory often requires a physical server with two to four gi-
gabytes of RAM. The resources of an ESX host, on the other hand, are shared among
all the virtual machines running on that host. Higher resource demands by one virtual
machine are offset by lower demands from another. Utilizing resource pooling between
ESX hosts further maximizes resource utilization.
Streamlining applications and operating systems enables further virtual infrastructure
savings. Tuning an application and operating system for a virtual environment has many
benefits. Examples of tuning include:
❖ Minimizing wasted resources (CPU, RAM, and disk)
❖ Generating warning and error logs when application and operating system is-
sues arise
❖ Minimizing the use of snapshots
❖ Reducing unneeded layered applications
❖ Scheduling antivirus software and guest OS–based backups to minimize con-
tention
❖ Using VMware Consolidated Backup or VMware Data Recovery instead of
guest OS–based backups, where appropriate
❖ Tuning code that accesses your databases to minimize contention by doing
more operations with fewer connections
❖ Minimizing the use of screen savers, including the pre-login screensaver
❖ Splitting data and log writes to different virtual disks and/or VMFS volumes to
reduce I/O contention and provide better DR capabilities
❖ Disabling unused services
❖ Disabling visual effects
❖ Minimizes use of defragmentation tools
❖ Avoiding mass scheduling of defragmentation tools, antivirus, or backup solu-
tions

Optimization / 77

VMware VMmark
VMware VMmark is a scalable benchmark tool for virtualized systems. This benchmark-
ing tool is used to quantify the performance of virtualized environments through the use
of diverse workload sets found in datacenters.
Traditional benchmarking tools focus on single workloads. In some cases, vendors have
tried to measure multiple workloads on the same platform but still use the benchmark-
ing tool in a one-to-one relationship with each workload being run. A virtual environ-
ment consists of multiple server instances all running different workloads on the same
hardware platform. Traditional benchmarks do not factor in load change among active
virtual machines, and they present the following challenges for a virtual environment:
❖ The benchmark specification must remain platform-neutral.
❖ The benchmark must capture the key performance characteristics found within
virtualized servers and the supporting infrastructure.
❖ The metric must be easy to understand.
❖ Benchmarks must provide a way to measure scalability that applies to small as
well as large servers.
The VMmark benchmarking tool measures the diverse virtual machine workloads and
generates an output that scales as resource capability and resource requirements change.
VMmark calls the work unit measuring these workloads a tile. A tile includes the col-
lection of virtual machines executing diverse workloads on one virtualization platform.
A system is benchmarked based on the total number of tiles that a physical system and
virtualization layer can accommodate.
Examples of tiles include a mail server, Web server, database server, file server, and ap-
plication server. These tiles are each sub-tests that are derived from standard load genera-
tors. Determining the maximum number of tiles that fit on one virtualization platform
provides a measure of its capabilities.

Figure 37: VMmark tiles
A typical VMmark benchmark test runs over several hours (at least three), with the ac-
tual workload metrics reported every 60 seconds. Typical trending benchmarks for scal-
ability run for at least one month.
VMmark is typically used by hardware vendors to compare results of the same hardware
platform to those of competitors. Results can be used for hardware selection based on
performance requirements (http://www.vmware.com/products/vmmark/results.html).

10. Business Resiliency

Business continuity is a strategic initiative in which the business owners determine and
outline requirements to ensure successful recovery in the event of a disaster. This in-
cludes determining what personnel and resources are necessary for the IT infrastructure
to survive should a disaster occur. The outcome is a business continuity plan.
Disaster recovery is a tactical initiative for an IT department to meet business continuity
requirements. As with business continuity, the goal is to determine who and what is nec-
essary for the IT infrastructure to survive a disaster. The outcome is a disaster recovery
plan.
Staffing and facilities do not depend on whether a solution is physical or virtual other
than the need for training and the facilities components to support the disaster recovery
requirements.
Depending on the level of interdependencies between systems and applications, it might
make sense to keep related applications in the same physical location. Both RTOs (re-
covery time objectives) and RPOs (recovery point objectives) need to be considered.
Consistency groups and dependency groups describe a collection of servers that perform
a single business function. For example, a consistency group could be made up of Web,
application, and database servers used to host an e-commerce front-end, where another
consistency group might be made up of an accounting file server and the application
server that relies on the data from it. It is important to plan your consolidation strategy
so that business functions are recoverable across applications rather than individual serv-
ers.

Redundancy
Component redundancy for each of the infrastructure and hardware components is
just as important in the virtual infrastructure as it is in a physical infrastructure. Multi-
pathing for storage, redundant network links, redundant hardware devices, and an N+1
server architecture are all methods for providing redundancy in the environment.

Backup and Recovery Strategies
Several data-protection strategies are available, including the use of guest OS–based
software, VMware Consolidated Backup, VMware Data Recovery, VMware snapshots,
storage snapshots, and storage replication. VMware Site Recovery Manager is a business
continuity and disaster recover process workflow product.

Business Resiliency / 79

Guest Operating System-Based Backup
Guest OS backups operate identically to physical machine backups. A backup agent is
installed within the virtual machines and is registered with a backup server. Recovery,
just as on a physical machine, can be done in either of two ways:
❖ Bare-metal restoration, if supported by the vendor and used for the backup
process, is one option. This is optimal for guest operating system–based back-
ups, since it does not require an OS installation.
❖ Installation of an OS and backup agent prior to performing restoration is the
other option.

VMware Consolidated Backup/vStorage API
VMware Consolidated Backup enables offloaded and impact-free backup for virtual
machines on an ESX host by allowing traditional file-based backup software to leverage
VMware virtual machine snapshot technology and efficient SAN-based data transfer.
Consolidated Backup is a backup enabler that allows for LAN-free backups handled by
a SAN-connected proxy server. The Consolidated Backup proxy server handles backup
processing, which frees resources on the virtual machines and the ESX host.
Consolidated Backup performs two types of backup: file-level backups and full virtual
machine method (file system level) backups. File-level backups are similar to traditional
backup methods in that full and incremental backups can be completed for Microsoft
Windows–based virtual machines. Full virtual machine backups are similar to bare-metal
restore backups and include all of the virtual disks, configuration files, NVRAM file, and
logfiles for a virtual machine. Full virtual machine backups are supported for all guest
OS types. The guest OS system state (such as registry information, system files, and boot
files) is not included when using Consolidated Backup–enabled backups.

Figure 38: Consolidated Backup infrastructure
Consolidated Backup requires installation on a Windows 2003 physical server. The
backup vendor client must be installed on this Consolidated Backup proxy server.

File-level Backups
Here are the high-level steps for file-level backups using third-party software with Con-
solidated Backup:
1. The Consolidated Backup server requests that a snapshot be added to a
virtual machine’s virtual disk.

80 / Business Resiliency

2. The block list for the virtual disk snapshot is provided to the Consolidated
Backup proxy server.
3. A driver on the Consolidated Backup proxy server mounts the block list.
4. The third-party backup software performs backup of the mounted virtual
disk.
5. The Consolidated Backup server requests that the snapshots be removed
from the base disks.

Full Virtual Machine Backups
Here are the high-level steps for completing a full virtual machine backup with third-
party software for Consolidated Backup:
1. The Consolidated Backup server requests that a snapshot be added to a
virtual machine’s disks.
2. The disk is exported to sparse format and stored on the Consolidated
Backup proxy server holding tank. The configuration file, NVRAM file,
and logfiles are copied to the same location of the holding tank.
3. The third-party backup software performs backup of the data in the
holding tank.
4. The Consolidated Backup server requests that the snapshot be removed
from the base disk.
Several components are involved in the Consolidated Backup–enabled backup process,
including the following:
❖ Hostd—A small application that runs on the ESX host and performs com-
mands on virtual machines on behalf of software such as VMware vCenter
Server and Consolidated Backup.
❖ VM to be backed up—VMware Tools is involved with the backup and is used
to make the virtual machines’ disks quiescent.
❖ Backup Proxy Server—This system contains the third-party backup software as
well as the Consolidated Backup framework.
❖ VMware Consolidated Backup framework—This consists of the generic com-
ponents:
❖ vcbMounter
❖ vLUN driver
❖ Common API across all supported backup software
❖ Integration module for specific third-party backup software

Full virtual-machine-method backups provide an option for disaster recovery. A full
virtual-machine-method backup can be set to export to a storage location presented to
the Consolidated Backup proxy server from a remote location. In the event of a disaster,
VMware Converter can be used at the remote site to import, register, and power on
the new virtual machine based on the original system. The RPO achieved is the time at
which the full virtual-machine-method backup occurred. This provides a bare-metal re-
covery mechanism that works well for both ESX and VMware Server, enabling small to
large businesses varying options based on budget and business continuity requirements.


VMware Data Recovery (vDR)
VMware Data Recovery is a virtual machine backup solution designed for smaller sites
and can support a limited number of virtual machines. vDR is fully integrated with
VMware vCenter and includes data de-duplication to save on disk storage for your full
virtual machine backups, and it offers file-level restore or entire images as needed. vDR
supports both full and incremental backups by utilizing the changed block tracking fea-
ture that the vStorage API provides.
vDR uses the vStorage API and supports the use of Volume Shadow Copy Service (VSS)
used by certain Windows operating systems. For virtual machines with Windows oper-
ating systems that do not support VSS, VMware Tools uses the default LGTO SYNC
driver. vDR will work with SAN or NAS devices. It also supports CIFS-based storage
such as Samba.
vDR supports de-duplication stores that are up to one terabyte in size. Each backup ap-
pliance supports the use of two de-duplication stores. There is no limit defined for the
number of de-duplication stores or the size of a store. However, when using more than
two stores or when the size of a store exceeds one terabyte, performance will be affected.

Virtual Machine Snapshots
Virtual machine snapshots take a point-in-time copy of a virtual machine, including
disk, RAM, and the virtual machine configuration. In the event of a guest OS failure,
an administrator can revert to one of the previously created snapshots.
An important consideration when utilizing snapshots is that Consolidated Backup
prefers that a virtual machine have no snapshots. It is also important to keep in mind
that a snapshot requires disk space to store a copy of the virtual disks and other informa-
tion from the snapshot. Snapshots can be used by setting variables to determine how
to proceed during the VCB operations. The variables are PREEXISTING_MOUNT-
POINT (removes a VM mount point used during a file-level VCB backup) and PREEX-
ISTING_VCB_SNAPSHOT (can be used to fail the VCB job or to delete the snapshot
before continuing, with “fail” as the default option).
For disaster recovery, snapshot technology can be used similarly to Consolidated Backup,
in that the snapshot can be placed on alternate storage locations. The snapshot can be
used to recover the virtual machine in the event of a disaster.
Snapshots can be used to assist in testing updates to the guest OS and applications run-
ning within a virtual machine. A snapshot captures the state of the virtual machine at
the moment the snapshot was taken. The snapshot data files subsequently collect any
changes made to the virtual machine since the initial snapshot was taken.
The snapshot includes the virtual disk, configuration information, and BIOS configura-
tion. The size of a snapshot can grow to be as large as the virtual disk it represents.
Several other files are created to support the snapshot created. X represents the number
of the snapshot taken in relation to previous snapshots:
❖ A file storing the state of the virtual machine when the snapshot was taken,
with the filename <VM name>-Snapshot<###>.vmsn
❖ A file storing the state of the virtual machine memory when the snapshot was
taken, with the filename <VM name>-Snapshot<###>.vmem


❖ A file acting as a write cache for changes to the virtual machine since the snap-
shot was taken, with the filename <VM name>-<nnnnnn>.vmdk
The Snapshot Manager within the VMware vCenter Server manages the snapshots as-
sociated with each virtual machine.
Snapshots allow you to test new software updates and provide a simple process to back
out of the change, if required.

Storage Replication
Replication of storage lies at the heart of any virtual infrastructure disaster recovery plan.
The relative ease of virtual server disaster recovery is enabled by the hardware indepen-
dence of virtual servers and the mobility created by encapsulating them in a small set of
files.
Planning and designing the storage and replication for a VMware vSphere infrastructure
is closely related to larger disaster recovery planning efforts. Several replication strategies
provide different levels of service and associated costs. For most organizations, a multi-
tiered replication strategy provides the most cost-effective VMware vSphere infrastruc-
ture replication solution. Organizations typically classify approximately 80% of their
servers and data with a 24-hour RTO and the remaining systems with stricter RTOs,
ranging from several hours down to continuous data protection (CDP) with synchro-
nous wide area replication.
Some companies with zero data loss (or as little data loss as possible) requirements use
data replication technologies such as synchronous replication to ensure that every write
is replicated before accepting the next write. Data protection technologies such as CDP
can also be used when faced with more comprehensive data recovery requirements.
As with any disaster recovery plan, the RPO and RTO are constrained by budget. VM-
ware vSphere enables extremely short RTOs, because servers can be restored and placed
in service very quickly. Indeed, it is often possible to have virtual machines up and
running within minutes following a disaster declaration. RPO, on the other hand, is
constrained by bandwidth. Replicating a large amount of changed data over a small data
pipe increases the RPO. This makes the selection of a storage replication strategy critical
to a successful VMware vSphere infrastructure disaster recovery plan.

storage replication strategies
Several replication strategies for VMware virtual infrastructures align nicely with backup
and recovery system tiers and technologies.
The 80% of servers and data that are backed up daily with a traditional 24-hour RPO
can be replicated through a variety of means from the main site to the recovery site.
This type of RPO is often best met by replicating backups. After the nightly backups
are completed (file- or vmdk-based), the data needs to be replicated to the recovery site
before the next backup window begins. When replicating a large volume of data, it can
become a challenge to maintain even a 24-hour RPO at a level of bandwidth that does
not exceed the budget.
The 20% of remaining systems can have much stricter RPOs, some even requiring con-
tinuous data protection and transactional replication. In these cases, replicating nightly


backups is insufficient. Either asynchronous or synchronous replication between storage
mediums is required.
Several strategies exist for enabling storage replication, depending upon RPO and bud-
get. The primary differentiation between these replication strategies is where they sit in
the storage stack.
Embedding storage intelligence directly into the storage fabric creates the highest perfor-
mance and transparency. Virtual SANs, LUN virtualization, remapping, and replication
can all be enabled transparently to the VMware Infrastructure. Synchronous remote rep-
lication of some or all of the storage traffic is accomplished by using a variety of storage
replication protocols.
When the higher-performing fabric-based replication solution is not practical or eco-
nomic, the recommended solution can be a man-in-the-middle replication server or ap-
pliance to provide asynchronous in-band or side-band replication of LUNs, and virtual
machine exports. These solutions include software- or appliance-based storage virtualiza-
tion or replication systems. They can also be used for asynchronous mirroring of operat-
ing system or vmdk-based backups on disk, virtual tape library, RDM LUNs, or NAS
storage. This replication strategy is still transparent to the VMware Infrastructure but
not to the storage infrastructure.
The man-in-the-middle replication scheme is most effective when combined with data
de-duplication solutions. Data de-duplication can dramatically reduce the amount of
replicated data, particularly when replicating backups that involve a great deal of dupli-
cate data. Adding data bandwidth compression further enables rapid replication with less
bandwidth required.
The last replication method involves inserting an operating system or application shim
to enable continuous data protection for individual applications. This is common in
highly transactional, business-critical applications, including financial, database, and
messaging servers. This method is very effective for application protection but needs to
be integrated with the application and operating system, creating a more complex solu-
tion.
Uncompressed, only about 14GB of data can be transmitted over a T1 line in a 24-hour
period. Determining the bandwidth required must be included in the RPO. The total
data, rate of data changes, and replication technology utilized must all be considered
when calculating replication time. RPOs can also vary widely depending upon the type
of server or application. When the necessary bandwidth is calculated, a cost-benefit anal-
ysis can be performed to determine whether bandwidth-saving techniques such as data
compression or data de-duplication are viable options for optimizing replication speed.

Fault Tolerance Technologies
vCenter Site Recovery Manager (SRM)
VMware Site Recovery Manager (SRM) provides end-to-end disaster recovery workflow
management and automation for a VMware Infrastructure. The workflow steps can be
automated to ensure that the recovery processes are consistently followed both for testing
and for real disaster recovery situations. SRM is built upon a storage replication founda-
tion.


Recovery plans are created within the tool to define the order of virtual machine failover
and startup. Automated, non-disruptive, disaster recovery testing allows validation test-
ing of the workflow in a fenced environment. This means that testing of the disaster
recovery workflow can be carried out without affecting the running production systems
that the workflow protects. This also minimizes testing costs and errors in manual steps
(such as breaking replication, rescanning LUNs, and re-registering the virtual machines
on the failover VMware Infrastructure environment).
The workflow acts as a working document of the recovery plans and provides the in-
structions for recovery. Frequent testing followed by any necessary adjustments to the
workflow results in higher success rates in the event of a true disaster recovery situation.
SRM provides sites that fall under regulatory compliance guidelines with a very cost-
effective method to define a workflow and complete testing. The testing process outputs
a report that can be used to support regulatory compliance auditor requirements.

Figure 39: VMware SRM inventory mappings

vCenter Heartbeat
VMware vCenter Server Heartbeat is designed to protect against failure of the OS or
hardware that the vCenter Server operates on. This protection occurs for both physical
and virtual vCenter servers through the following protection levels: server protection,
network protection, application protection, performance protection, and data protec-
tion. This operates in a passive clustering model. Should any of the protection levels fail
on the primary server, a secondary server takes over.
VMware vCenter Heartbeat provides high availability and disaster recovery for VMware
vCenter Server and related components such as licensing and plug-ins. vCenter Server
Heartbeat enables monitoring, replication, and rollback for instances of vCenter Server
enabling effective disaster recovery in the event of loss of the primary datacenter. The
secondary server uses the same name, same network address configuration, and the same
file and data structure and can run the same applications/services as the primary server.
The secondary server is activated if a dedicated network connection between the two
loses a heartbeat. The heartbeat is maintained using keep-alive messages. When the sec-
ondary server detects the loss of a heartbeat, then a packet filtering mechanism changes
state to allow the secondary server to communicate with the primary node network con-
figuration.


VMware Fault Tolerance
VMware Fault Tolerance eliminates downtime due to hardware failures, by creating a
live shadow virtual machine to run on a second ESX host, receiving and executing each
memory operation and each CPU instruction in virtual lockstep, using vLockstep tech-
nology, with the primary VM. If the primary host and protected VM fail, processing is
seamlessly cut over to the secondary VM running on an alternative ESX host without
requiring a reboot. A new shadow VM is then automatically configured to run on anoth-
er ESX host within the cluster, which then runs in virtual lockstep with the new primary
VM. Fault Tolerance is a simple, selectable feature that can be implemented for specific
VMs deemed to be critical to the operation. Both primary and secondary VMs utilize
shared storage and can be VMotioned among ESX hosts. Fault Tolerance eliminates the
need to purchase and manage traditional clustering offerings dependent upon stand-by
hardware.

Figure 40: VMware FT provides zero downtime of applications and OS

VMware High Availability
VMware High Availability provides quick and automated failover of applications in the
event of ESX host failure. When a host fails, HA quickly detects the failure, sorts all the
VMs on the host based on their restart priority, and then places them on alternative ESX
hosts within the cluster. The virtual machines reboot very quickly compared to physical
servers, although the in-memory state of the virtual machines at the time of host failure
is lost.

Figure 41: VMware HA protects and restarts VMs from host failure or VM failure


Networking Strategies for Disaster Recovery
While storage replication is the foundation of VMware Infrastructure disaster recovery,
enabling effective replicated networking is a requirement that is often overlooked. In
the event of a disaster declaration, users may not be able to get access to their personal
computer and may only be able to work from another office or remote location. Both
application and network access strategies are essential to enabling quick and effective
recovery.

Thin Remote Application Access
A crucial piece of the business continuity plan is to re-establish user access along with
server recovery. Architecting a datacenter failover that is transparent to application users
requires that the primary site network scheme be moved to the failover site along with
the applications and processing. Providing thin network-based application delivery using
common methods such as application virtualization or streaming Web-based applica-
tions, or connectivity to remote virtual or thin desktops through a centralized applica-
tion gateway hub or portal, enables continued access to recovered applications without
reconfiguring client access.

Network Infrastructure Extension
When some form of remote application access is available, networking name and address
spaces must be transferred from the primary site to the disaster recovery site to allow the
systems to resume activity without application or network reconfiguration. When possi-
ble, the primary network is virtually extended not only into the VMware Infrastructure,
but also out to the remote recovery site using CWDM or MPLS WAN technologies,
creating an extended virtual grid or clustered datacenters. Network and port address
translation schemes are used to protect the internal address space.

Datacenter Access Cutover
When the network infrastructure and applications are replicated and recovered, a
method must be available to cut over application access from the primary datacenter to
one of the disaster recovery datacenters within the RTO without reconfiguring clients.
The most common way of achieving this is through the use of dynamic DNS to change
address resolution for the application delivery points, redirecting user access to the repli-
cated and restored disaster recovery site. Another way is through the use of network load
balancers such as those from Cisco or F5 Networks.
The use of consolidated application delivery points simplifies the redirection of users,
requiring only redirection of DNS resolution to restore application access. This consoli-
dation and redirection can occur easily with application portals or hubs for Web access,
secure SSL VPN gateway products for streamed or virtualized applications, or virtual
desktop connection brokers for organizations hosting virtual desktops in the datacenter.

Security Considerations for Disaster Recovery
Many virtualization deployments must meet specific regulatory compliance guidelines
for security. Consider any firewalls, packet filters, or intrusion detection systems that can
affect the virtual machine after it is consolidated. For example, a physical machine that is


not protected by a firewall might end up on an ESX host that is behind a firewall. Also,
IDS systems might not be prepared for the types of traffic that are seen coming from
new virtual machines on a network segment.
Supportability is often a customer concern. Some software vendors say that they will
not support their applications within a virtual machine. If this is the case, the customer
should contact the ISV directly and state the need for support. If future license revenues
are at stake, the ISV might be more inclined to provide some sort of support statement.
In the past, VMware has participated in these sorts of conversations. See VMware’s part-
ner catalog Web page for details on existing agreements with various ISVs.
Using VMware vCenter Server roles and permissions is a best practice to support regula-
tory compliance and change management requirements. Changes made by individuals
are logged for use by an audit team.
Creation of security pods is one method used by customers to satisfy audit requirements.
A specific grouping of virtual machines using the same VMFS volumes and virtual
switches is an example of a security pod. Add specific roles and permissions to complete
the setup. Remember that the roles and permissions are tied to the accounts on the VM-
ware vCenter Server.

11. Security

Security Considerations
As with physical infrastructures, virtual infrastructure architecture needs to reflect
security considerations at every juncture to ensure an effective, useful, and secure envi-
ronment. In many cases, specific regulatory guidelines must be met. VMware vSphere
incorporates an ability to build a virtual infrastructure that not only addresses virtualiza-
tion-specific security concerns but is more secure than the physical alternative.
As with physical infrastructure, security design must incorporate firewalls, packet fil-
ters, and intrusion detection systems. The flexibility of virtualization, though, requires
different considerations. For example, a virtual machine protected by a firewall might
VMotion to an unprotected ESX host. Or an Intrusion Detection System (IDS) may be
unprepared for traffic coming from new virtual machines on a network segment.
Some software manufacturers refuse to support their applications within a virtual ma-
chine, sometimes claiming a lack of security as justification. A properly architected vir-
tual infrastructure leveraging the capabilities of vSphere enables an environment at least
as secure as the physical equivalent. A customer can often contact the ISV and obtain
support—VMware has participated in these types of discussions. See VMware’s partner
catalog Web page for details on existing agreements with various ISVs. If no policy is in
place, consider working with VMware to get the ISV to establish a reasonable support
policy.
Utilizing VMware vCenter Server roles and permissions is a best practice to support
regulatory compliance and change management requirements. Changes made by indi-
viduals are logged for use by an audit team.
Creation of security pods is a method to satisfy audit requirements. A specific grouping
of virtual machines using the same VMFS volumes and virtual switches is an example of
a security pod. Add specific roles and permissions to complete the setup. Roles and per-
missions are tied to the accounts on the VMware vCenter Server. Other considerations
are the type of hypervisor used in the virtualized infrastructure. Both ESX and ESXi
are bare-metal hypervisors and are fully supported, with each having its own benefits.
VMware is moving towards an ESXi-only solution, where security risks are reduced due
to a smaller attack surface. Where ESX contains a modified version of Red Hat Linux,
in VMware terms called the Service Console, ESXi does not. The functionality of the
Service Console, however, has been moved to an appliance which is optional. Currently
many of the patches released are Service Console related. Not only will using ESXi de-
crease security risks, it will also simplify patch management and reduce all related opera-
tional costs.

Security / 89

Enhancements
VMware vSphere 4.0 introduces many new security-related enhancements. These
enhancements include memory hardening, kernel module integrity, and extended
permissions for vCenter. As introduced in Chapter 2, VMware vSphere also includes
enhanced security capabilities with VMsafe, vShield Zones, and the VMware vNetwork
Distributed Switch.
Memory hardening can best be described as a method for storing drivers, libraries, and
applications in random memory locations. This location is non-predictable, which
makes the hypervisor less vulnerable to memory exploits and malicious code.
Authenticity of modules, drivers, and applications is assured by a digital signature. This
signature will be verified when the module, driver, or application is loaded by the VMk-
ernel.
One of the areas where vCenter permissions have been vastly improved is datastore
privileges. Datastore privileges enable you to control who has access to a datastore and
what kind of operations they are allowed. A good example of a new option is “allocate
space.” With “allocate space” one can control who can create new virtual machines,
snapshots, or clone or create a new virtual disk on a specific datastore.
Privileges can also be controlled for vNetwork Distributed Switches and vNetwork
Distributed Port Groups. With Distributed vSwitches, setting the correct permissions
has become much more important because of the risk and impact changing a
Distributed vSwitch or Port Group has.

VMsafe
VMware VMsafe is a new security technology for VMware environments. VMware VM-
safe enables, through published APIs, third-party vendors to provide virtual appliances
that automatically protect everything running within the vSphere virtual machines, thus
no longer requiring the deployment of individual agents. Virtual security can be as-
signed specifically to individual VMs. This makes it possible to monitor within a virtual
machine and stop the execution of, for example, a virus. Other threats/risks that can
easily be identified are rootkits and malware, and VMsafe even monitors for disallowed
process execution.

vShield Zones
VMware vShield Zones is an application-aware firewall technology that is built for vir-
tualization and uses VMsafe APIs. vShield Zones builds a firewall “fence” around a VM
cluster by creating security zones internal to the virtual machine that follow it as it mi-
grates between hosts. This ability enhances security by providing firewall appliance func-
tionality without the complexity of network management. The following figure shows
an example of securing specific zones in your virtual network with vShield Zones.

90 / Security

Figure 42: vShield Zones

vNetwork Distributed Switch (vDS)
The vDS enables monitoring of traffic over a virtual switch, thereby resolving one of the
largest security concerns regarding virtualization. It enables application of policies such
as availability, encryption, and maximum latency to a virtual machine that then follow
the virtual machine as it migrates between ESX hosts. A VMware vDS also provides an
open API, thus allowing third-party manufacturers to provide their own switch plug-ins.
The Cisco Nexus 1000V virtual switch, for example, provides a full suite of Cisco secu-
rity features, including access control lists, private VLANs, IP Sourceguard, Netflow, etc.

12. VMware View

The benefits of a virtual infrastructure are well documented. Consolidation and stan-
dardization of resources enables greater physical and human resource efficiency. Virtual
machine encapsulation provides mobility, enabling replication and recovery of logical
servers in ways unheard of in a physical datacenter. High availability, distributed re-
source scheduling, and consolidated backup services add additional value to virtualized
servers.
Organizations implementing comprehensive VMware-based virtualized datacenters
often find a divide between the manageability and simplicity in their virtual server infra-
structures and the cumbersome distributed management of traditional physical desktop
computers. This separation between virtualized servers and physical desktop comput-
ers becomes problematic when planning for disaster recovery—a replicated, recovered
virtual server infrastructure without access to the desktop computers cannot provide
complete business continuity. Early adopters of VMware looked for a way to extend
the scope and benefits of their VMware Virtual Infrastructure to include the client-side
computing resources to provide a consistent and complete virtualized infrastructure,
and VMware responded by launching a new industry—Virtual Desktop Infrastructure
(VDI).
VMware View delivers desktops as a managed service as part of the virtualized infra-
structure. A VMware View VDI solution is built by accessing ESX-hosted virtual desk-
top images using a high-performance PCoIP remote display protocol through the View
Manager virtual desktop connection broker. Individual desktops can be assigned to each
user or allocated from a shared pool of desktops: persistent pools for logging users into
the same desktop each time and non-persistent pools for generic desktop access. VM-
ware View can suspend virtual desktops not in use to free up compute resources and to
save power. Additional components manage printing, client hardware, application distri-
bution, and patching services.
Combining View and vSphere enables the server, desktop, disaster recovery, network,
and storage computing infrastructures to be deployed, managed, replicated, and recov-
ered as a unified whole throughout the enterprise. The centralization of management,
administration, and resources significantly lowers costs, as shown in the ROI section of
this book. Security is enhanced as the desktops move to the datacenter where they are
backed up, managed, and replicated off-site for disaster recovery purposes. The desktop
computing experience improves through ubiquitous access from any device anywhere a
user can get a browser. And business agility and productivity increase through rapid pro-
visioning and a reduction in hardware-related user downtime.

92 / VMware View

Among the PCoIP enhancements included are multimedia redirection, dynamic host- or
client-side media rendering, dynamic Adobe Flash control, USB redirection for seamless
peripheral support, dynamic bandwidth adjustments and network optimizations, dy-
namic audio adjustment and audio redirection support for audio peripherals, and multi-
monitor support for up to four monitors, along with 1920x1200 resolution per display.

VMware View Product Suite
Whereas VMware formerly referred to its desktop virtualization products as VDI, they
are now part of the VMware View portfolio. The standard edition of View enables VDI
with View Manager, while the Premier Edition enables storage optimization with View
Composer, application virtualization with ThinApp, and client virtualization with Of-
fline Desktop. Figure 43 represents a comprehensive VMware Infrastructure environ-
ment, including the VMware View product suite components.

Figure 43: VMware View VDI infrastructure

View Manager
View Manager, formerly called Virtual Desktop Manager, provides a single console to
provision, deploy, and manage virtual desktops and includes View Connection Server,
View Agent, View Client, View Portal, View Administrator, View Printing, and Unified
Access.
❖ View Connection Server works with vCenter Server to provide advanced man-
agement capabilities. It installs on a Microsoft Windows server that is part of
an Active Directory domain.
❖ View Agent runs on each virtual desktop and enables session management and
single sign-on. View Agent can be installed on a virtual machine template, en-
abling automatic inclusion for desktops created from the template.
❖ View Client runs on a Windows PC or a thin client and connects users to their
virtual desktops through the View Connection Server.

VMware View / 93

❖ View Portal provides alternative Web browser access to virtual desktops.
❖ View Administrator provides a Web-based console for administrators to man-
age virtual desktops, including configuration and policy settings.
❖ View Printing is installed on the virtual machines and configures them for
printing. It receives, decompresses, and decrypts print data, then converts the
common data format into printer-specific formats for sending to the print
device. It automatically makes all necessary printers available while eliminat-
ing the need to install and maintain printer drivers on virtual desktops. The
advanced print stream compression can reduce network utilization by 98% and
enables high quality printing even over WAN connections.
❖ Unified Access leverages View Manager’s secure connection brokering for
other platforms accessible by RDP, including terminal servers, blade PCs, and
physical PCs. It enables load balancing of multiple terminal servers along with
monitoring and auditing from within View Manager.

View Composer
VMware View Composer provides image management for environments with large seg-
ments of homogeneous desktops. It is a complementary solution to traditional manage-
ment technologies, and while it can dramatically reduce storage, it also provides a new
method of management and control.
View Composer uses VMware Linked Clone technology to rapidly provision desktops.
A linked clone shares virtual disks with a master virtual machine image, enabling admin-
istrative tasks such as patching to take place at the master image level. It also can reduce
storage requirements. The clones can be powered on, suspended, snapshot, and recon-
figured independently of the parent. User profiles are separated from the desktop image
and stored as user personality disks, enabling independent administration.
View Composer image management provides the ability to add more storage as required
and to retire an existing storage array with three primary techniques: refresh, recompose,
and re-balance.
❖ Refresh enables resetting a linked clone desktop back to the state of initial roll
out.
❖ Recompose alters a deployed desktop state such as assigning it a snapshot of a
different master virtual machine.
❖ Re-balance balances virtual machine disks across available datastores to enable
efficient usage of available storage.

VMware ThinApp
VMware ThinApp is an agentless application virtualization solution that decouples the
application from the OS. It encapsulates each application and all components required
for it to run in an isolated container. Applications can be packaged once and deployed to
multiple devices across both virtual and physical systems, minimizing the desktop im-
ages to be managed. User-specific configurations and changes are stored separately.
VMware ThinApp enables VDI administrators to reuse virtual machine templates and
reduce storage requirements by storing and streaming application and user-specific
configuration data from a central server. A pool of users can utilize the same central

94 / VMware View

application. Multiple versions of the same application can run side by side without con-
flict, because resources are unique to each application. ThinApp applications can also be
linked together to share common independent components (e.g., Java, .Net), minimiz-
ing package size. Application patches and updates are streamlined and can take place
while applications are in use.

Offline Desktop—Experimental Use
Remote connectivity is increasingly becoming ubiquitous as even airlines are beginning
to outfit their planes with WiFi. For users requiring an ability to work while discon-
nected, Offline Desktop allows the movement of complete virtual desktops between the
datacenter and the laptop or other physical devices, with administrator-defined security
levels and encryption policies intact. Offline Desktop enables designated end-users to re-
locate their View virtual machines to a local physical computer and back. The processes
are termed check out and check in, respectively.
❖ A check-out operation is initiated by the client and ends with a local copy of
the online desktop on the user’s local computer. A user can check out after
connecting and authenticating to View Manager at least once, when a policy
allows offline access and when the desktop is not currently checked out.
❖ A check-in operation is initiated by the client to push changes made to a previ-
ously checked-out desktop up to the datacenter. A user can check in when con-
nected and authenticated to View Manager, assuming an open offline session
exists on the client device and the desktop retains policy permission for offline
usage.
When checked out, the virtual machine has a “heartbeat” back to the datacenter allow-
ing administrators to deactivate it if necessary. When the user reconnects, only the delta
is checked in. The Offline Desktop must be managed by vCenter Server and must either
be an individual desktop or in a persistent desktop pool. ThinApp applications stored on
a shared network drive will not be checked out.
The administrator can optionally encrypt communication and data transfer between the
offline desktop and View Connection Server, although the data on an offline desktop
will always be encrypted.
Policies are used to assert control over product components at the global, desktop pool,
and desktop user level. By default, each user-level policy inherits its setting from a pool-
level policy that, in turn, inherits its setting from a global policy. Policies include the
ability to limit the amount of time an offline desktop can run without successfully con-
tacting the View Manager server for policy updates.

VMware View Infrastructure—Design Considerations
The virtual desktop workload is very different from a virtual server workload; a View
implementation can create hundreds to thousands of virtual machines, each with a
smaller resource footprint than a virtual server. Proper sizing is critical to ensure both
adequate capacity and performance, as the end-user experience will make or break a VDI
project.
The first step is to assess the physical desktop usage. VMware Capacity Planner along
with tools such as Perfmon or LiquidWare Labs Stratosphere can be used to measure

VMware View / 95

desktop performance, including CPU usage, memory consumption, storage throughput,
storage IOPS, and network throughput. The results will assist in designing an optimal
VMware View infrastructure. Of course, a trade-off exists between performance and re-
source dedication. VDI technology is advancing at a rapid rate and sizing considerations
are similarly rapidly changing.

Architecture Planning
In planning the View architecture, we recommend taking a conservative approach that
errs on the side of excess performance and capacity. It is easier to increase consolidation
densities later than to salvage a project that has garnered a reputation for poor perfor-
mance. The environment should be designed to accommodate usage spikes. It is impor-
tant for the storage and network groups to participate along with the server group and
desktop team in the planning.
An important architectural consideration is whether or not to run View on the same
ESX instances as the server virtual machines. VMware recommendations are generally
to use a set of ESX servers in a cluster dedicated to desktops, although utilizing vCenter
Server to manage both environments. A large environment will likely have multiple clus-
ters for both View and servers.
Some other architectural considerations include:
❖ Protocol: FC, iSCSI, NFS. If using VMware View Composer, NFS may be a
more desirable protocol, as it is not restricted by the VMFS limitation of an
eight-host maximum.
❖ Storage: Type of shared storage and drives and whether or not to utilize a tiered
solution (e.g., FC for system drives, NAS for data drives). Special consider-
ations also need to be taken into account when planning storage for virtual
desktop infrastructures. Virtual desktop images are generally much smaller
than server images, creating a larger vmdk-to-physical disk or LUN ratio. VDI
storage architecture and management practices need to deal with concurrent
access to thousands of vmdk files. Resource requirements must be factored
into activities such as desktop search tools. Storage should be sized for peak
usage, not average. Enough IOPS and bandwidth must be available or perfor-
mance will suffer. The storage footprint, however, can be minimized with View
Composer, virtualizing the applications with ThinApp, utilizing thin provision
storage, using nLite and vLite to streamline the guest OS, and with array-based
snapshots. Activities such as anti-virus scans, OS patching, and file updates
should be staggered and deployed at the least busy times.
❖ Servers: Considerations include whether to use traditional rack-mount serv-
ers, blade servers, or new, specifically designed virtual infrastructure hosting
platforms such as the Cisco Unified Computing System. When sizing, it is im-
portant to plan for virtual machine memory overhead—84MB for each 32-bit
VM with 1GB RAM. Page sharing should be maximized. Over-commit should
be effectively leveraged and ESX host-level memory swapping avoided.
❖ Network bandwidth: Just as in storage systems, activities such as anti-virus
scans, OS patching, and file updates should be staggered and deployed at the
least busy times.

96 / VMware View

❖ WAN bandwidth: When accessing virtual desktops over a WAN, extraneous
services and graphical add-ons such as themes and screen savers should be dis-
abled. Bandwidth requirements depend upon the display protocol being uti-
lized (PCoIP or RDP), the usage cases, and the amount and type of printing. A
good rule of thumb is to allow at least 128K per user while keeping latency at
less than 150 ms.
❖ View Composer: The heterogeneity of the environment is important in deter-
mining whether or not to use Linked Clones. View Composer supports only
Windows desktop images. Similar or superior storage consolidation can be
achieved with SAN array products such as NetApp FlexClones or EMC Snap.
Today the clones must be manually registered in View Manager, and automatic
pool sizing cannot be utilized; however, storage array manufacturers are work-
ing to integrate their products into View Composer in order to provide en-
hanced functionality and scalability.

View Composer/Linked Clones Deployment Considerations
❖ Keep the number of VMs per LUN under 64 for best performance.
❖ Each desktop can only have one user data disk.
❖ Store the OS disk and the user data disk on different data stores.
❖ Keep the system disk from growing too big. Updates directly to a linked clone
system disk can potentially cause inefficiency by increasing their size.
❖ A SAN is required to support View Composer in a cluster; it will not work on
a local disk. View Composer supports a maximum of eight hosts in a cluster
unless running NFS as noted above.
❖ Be conservative with storage over commitment.
❖ Available storage needs to be managed regularly by the administrator, with used
space kept under 95%, as exceeding this level can result in loss of performance.

Capacity Planning
Capacity planning is difficult but critical. It is facilitated by evaluating the results from
the physical desktop environment assessment. Typical desktop workloads for CPU,
memory, network, and disk requirements can be modeled for different categories of us-
ers, but should be validated against resource oversubscription. Alternatively, the View
environment might be sized according to worst-case performance requirements.
❖ Estimate the CPU requirements. VMware studies show approximately eight
typical knowledge workers per virtual desktop per core, although a 2009
in-depth study of evaluating the scalability of VMware View on a Cisco Uni-
fied Computing System (UCS) showed an average of 20 users per core (two
4-core Intel 5500 series 2.93 GHz CPUs per UCS blade and 96 GB RAM).
See http://www.emc.com/collateral/software/white-papers/view-ucs-vmax-
whitepaper.pdf.
❖ Estimate the memory requirements. The ability of vSphere to utilize page shar-
ing significantly reduces the amount of memory normally required when aggre-
gating virtual desktops. VMware recommends provisioning a minimum of 512

VMware View / 97

MB RAM for a Windows XP virtual machine without memory sharing, or half
that with memory sharing.
❖ Estimate the network requirements. In addition to examining the physical net-
work usage patterns, virtualization considerations include the remote display
protocol being used, the printing requirements, shared/redirected folders, and
multimedia requirements.
❖ Estimate storage capacity requirements. A formula guideline is (size of vmdk) +
(VM RAM) + (VM suspend/resume) + 100MB per VM for logs.
❖ Estimate storage performance requirements. Storage performance is the most
difficult and important element to design correctly. In addition to examining
the physical usage patterns, virtualization considerations include whether other
systems or virtual machines are sharing the same spindles and the VMware
ESX disk I/O. Also important are usage patterns such as boot periods, user ini-
tiated desktop search, defrag, virus scan times, etc.

Guest OS Sizing Considerations
Unnecessary services, device drivers, and add-ons, including graphical screen savers,
should be disabled. Desktop images should be eliminated by using centralized file stor-
age, roaming/virtualized profiles, de-coupled applications, and redirected application
data, cookies, favorites, and templates.

VMware View Infrastructure—Deployment Considerations
Careful piloting is essential for a successful enterprise View deployment. It is important
to simulate the anticipated production VDI workloads as accurately as possible and then
to evaluate the performance in actual, though limited, usage scenarios. This will either
validate the capacity planning or allow a reevaluation of resource requirements.
Another important area when considering VDI is management of user expectations and
change management. Users who are assigned virtual desktops should know about policy
changes such as capped storage, no administrator privileges, and so on. The VDI pilot
will assist in enabling a smooth transition to View from the physical desktop.
A pilot program of 25 to 100 virtual desktop users is typically adequate to test-drive
the policy decisions and better understand user requirements. During the pilot, group
policies can be adjusted and finalized. User satisfaction surveys can identify productivity
issues. The virtual desktop templates (virtual machine image) can be finalized for use in
the full-scale implementation.
It is important to understand that moving an organization from physical desktops to vir-
tual desktops is as much a culture change as a technology change. If managed properly, a
View VDI implementation can save your organization considerable capital, operational,
and management expense, as well as enabling increased user productivity. A View VDI
deployment puts your desktop assets in a position to benefit from all the current and
future innovations that virtualization has to offer.

Appendix A. Virtualization Technologies

Virtualization technology1 has a long history. We’ll describe some different virtualization
technologies, compare these approaches, and provide a context for VMware virtualiza-
tion. We’ll describe only system-level virtualization, as opposed to process-level virtual
machines such as Java virtual machines.
Virtualization was first developed in the 1960s and later popularized on IBM main-
frames in the 1970s. The most famous version implementation was the VM/370. These
systems partitioned hardware on a single computer into virtual machines. The goal was
to enable mainframe computers to perform different tasks at the same time. Mainframe
computers were expensive; virtual machines maximized their resource utilization by run-
ning concurrent tasks for different users.
UNIX vendors adopted hardware-level virtualization and partitioning technologies, and
then later adopted software virtualization.

Operating System Virtualization
Operating system virtualization splits the operating system of a single computer into
multiple partitions, which can run multiple instances of the same operating system. Ex-
amples of this include chroot jails in UNIX.
Logical partitioning, also known as LPAR, is found in mainframe computers such as
IBM System z (and less commonly on other IBM systems), as well as on computer sys-
tems from other vendors. In logical partitioning the resources of a physical computer are
partitioned so the computer’s memory may be split, allocating a specific range to each
partition. Hardware assistance is often used to partition a system but is not necessary for
operating system virtualization in general.

Hardware Virtualization
In this model, which VMware technology uses, one or more abstractions of a computer
are created. This enables more flexibility, since it allows one computer to run several dif-
ferent operating systems.
Hardware virtualization can be performed using two different methods: hosted or hyper-
visor. Examples of VMware products in each category are given.

1. Terminology for virtualization technologies varies by vendor. Where different terminologies can be
used, we have defaulted to using the VMware terminology in this document. Concepts included can be
applied to other virtualization implementations.

100 / Virtualization Technologies

Hosted
Hosted virtualization (VMware Workstation, VMware Player, VMware ACE, VMware-
Server, and VMware Fusion) relies on having a standard operating system between the
physical computer and the virtualization layer. This requires installation of an operating
system such as Microsoft Windows, and then installation virtualization software such as
VMware Workstation on top of it. Finally, a guest operating system such as Windows or
Linux is installed in one or more virtual machines running within VMware Workstation.
The hosted virtualization platform depends upon the host operating system for resources
and is also impacted by any issues that might affect the host operating system. If the host
operating system gets busy or if it crashes, the virtual machines are affected. A benefit
is that hosted virtualization systems can run on computers that support common OSes
such as Microsoft Windows, increasing compatibility.

Hypervisor or Bare Metal
Hypervisor virtualization (VMware ESX, VMware ESXi, part of VMware vSphere)
platforms have a partitioning layer that runs directly on top of the hardware and below
higher-level virtualization services that provide a virtual machine abstraction. The hy-
pervisor is installed on the computer, just as though it is an operating system. It provides
the capability to create virtual machine partitions, with a virtual machine monitor run-
ning within each partition.
Hypervisor virtualization platforms eliminate the overhead of typical operating systems
and have direct access to and control of the actual hardware resources. The performance
of virtual machines operating on top of bare metal virtualization is closer to native per-
formance than a hosted approach.

Virtual Machine Monitor
The VMM is a layer of software that runs between the hypervisor or host operating sys-
tem and a virtual machine. It manages the resources and their allocation to the virtual
machines running on the system.
The VMM decouples the software from the hardware underneath. As a famous quote
from computer scientist David Wheeler states, “All problems in computer science can be
solved by another level of indirection.” VMM’s decoupling capability provides substan-
tial control over how the guest operating system accesses the hardware.
VMM may be implemented in many ways. Some computer architectures in the past
were designed to be virtualized, but many CPUs, including the x86 family (except for
recent editions), are not, thus requiring techniques such as binary translation to work
around this limitation.
VMMs have the primary task of executing instructions on the virtual CPU and emulat-
ing virtual devices.

CPU Virtualization
The VMM can follow one of several techniques for CPU virtualization. These examples
are specific to x86 virtualization.

Virtualization Technologies / 101

Full Virtualization
With full virtualization (also known as native virtualization), the guest OS is presented
with a virtual hardware abstraction that represents a physical machine. This does not
mean that the virtual machine is identical to the underlying hardware. A virtual ma-
chine can be thought of as a VMware-brand PC that is standardized to the VMware
Infrastructure architecture but is different from the underlying hardware. The virtual
machine is recognized and accessible to the operating system or applications software
just as if it were a physical machine, so no modification to the software is necessary. VM-
ware has traditionally used a binary translator to overcome some limitations in the x86
architecture that made virtualization difficult. Note that hardware-assisted virtualization
is a variant of this, where x86 hardware designed for virtualization assists in this task. It
is important to remember that in the case of full virtualization, a standard operating sys-
tem such as Windows or Linux, without modifications, will run in a virtual machine.

Para-Virtualization
With para-virtualization (also known as OS-assisted virtualization), the guest OS is
presented with a modified hardware abstraction. This requires operating systems to be
modified and ported to this particular virtualization platform. This reduces operating
system compatibility, but that is a trade-off against potential increases in performance of
certain CPU-bound applications that run on systems without virtualization hardware.
This performance increase is achieved by hypercalls, a communication method that oc-
curs between the guest OS and the hypervisor, but the performance advantage can vary
greatly, depending on the workload. However, each guest operating system, such as
Linux, needs to be modified. VMware has traditionally offered full virtualization, but
aspects of para-virtualization have been offered as an option for enhanced device drivers
that increase the efficiency of guest operating systems.

Hardware-Assisted Virtualization
Recent CPUs from Intel (Intel VT) and AMD (AMD-V) implement hardware assis-
tance for CPU virtualization; the first generation of these CPUs was released in 2005
and 2006. This method overcomes some of the problems in x86 virtualization that origi-
nally led companies such as VMware to pursue full virtualization with a binary transla-
tor. Although the binary translator can outperform first-generation hardware-assisted
virtualization, future enhancements are expected to improve performance and flexibility
in the programming model. Hardware-assisted virtualization can be considered a special
aid to enable virtualization, and it gives the x86 architecture some capabilities of main-
frame CPUs that were lacking in the original x86 CPUs. After all, most people didn’t
expect x86 computers to be powerful enough to run multiple operating systems at once,
but now they are capable of running many virtual machines at once.

Device Virtualization
Note that CPU virtualization is not sufficient to create a fully functional virtual ma-
chine. VMM provides many other critical components, such as the computer’s memory
(memory management unit, also known as MMU), devices, and I/O, which are re-
quired for a fully functional x86 computer. The complexity of creating fully virtualized

102 / Virtualization Technologies

memory, devices, and I/O subsystems can be as great as the effort required for core CPU
utilization itself. Hardware virtualization virtualizes the underlying hardware as virtual
devices and presents them to the guest operating systems, and the virtual hardware pre-
sented is consistent for all virtual machines regardless of the underlying physical hard-
ware. This means that a virtual machine running on one hardware platform can easily be
moved to another platform, as the virtual devices are the same.
For example, let us take a Brand A computer, installed with virtual machine software
and configured with Linux to run within a virtual machine. That instance of Linux
is not configured for the Brand A computer (in the device drivers, size of disk, etc.).
Instead, it is configured to run against the configuration of the virtual machine (a “VM-
ware brand PC”). If you choose to replace the Brand A computer with Brand B, you
simply move the virtual machine (which is just a set of files) to the Brand B computer,
and the instance of Linux you installed earlier will run without any need to reconfigure
its drivers. This greatly simplifies the difficulties associated with hardware upgrades.
Furthermore, VMware Infrastructure also provides VMotion capability, which migrates
a running virtual machine from one computer to another (provided they share the same
storage where the virtual machine resides) without downtime.
Here are some specific examples. In a VMware virtualization environment, network
adapters are presented as AMD PCNet devices, storage devices are presented as SCSI
(even if the underlying physical devices are SAN, iSCSI, or SATA devices), and CPUs
are presented as the underlying CPU architecture type. Other computer devices such as
DVD/CD-ROM drives and floppy drives are presented using either physical devices or
device images as the source.

Other Forms of Virtualization
There are other methods of virtualization not directly related to VMware Infrastructure.
Some of the commonly used terms for these methods are given below.

Emulation
A virtual machine simulates the hardware needed in a way that allows it to run on a plat-
form with a different CPU than it was originally designed to work with, e.g., PearPC,
a PowerPC platform emulator; Microsoft Virtual PC, a 32-bit x86 emulator for Apple
PowerPC Macintosh; Bochs, a 32-bit x86 emulator project.

Storage Virtualization
The process of abstracting a logical storage device from a physical device. These are
found in many areas, from virtual disks in VMware products to Fibre Channel or IP
network storage devices such as IBM’s SAN Volume Controller (SVC), EMC Invista, or
LeftHand Networks Virtual SAN Appliance (VSA). The physical location of the storage
can be on a SAN, but the representation might be iSCSI. The software handles the map-
ping between the physical storage and the logical storage.

Virtualization Technologies / 103

Network Virtualization
VLANs (virtual LANs) are used to segment networks on physical networks. This is dif-
ferent from the virtualized network devices available in VMware, although these two
technologies can coexist.

Summary
This has been a quick overview of virtualization, spanning virtual machines, virtualiza-
tion technologies (software- and hardware-level virtualization), approaches to CPU vir-
tualization (full, para-, and hardware-assisted), what virtual machine monitors provide,
and four key properties of virtual machines.
In your daily work and conversations, it’s unlikely that these terms will come up, but
understanding these concepts might become useful if you ever get into discussions where
some parties are confused about basic concepts in virtualization. A four-way server
built with dual-core CPUs and 32GB RAM, for example, is going to be used as a vir-
tualization host running VMware ESX. A beginner’s misconception might be that the
four-CPU system will be logically partitioned into four virtual machines so that each
partition is a two-core CPU, and you partition the RAM into 8GB partitions each. This
type of misconception can stem from knowledge of LPARs, but you know that hard-
ware-level virtualization is quite different from the flexible software-level virtualization,
where you can create a wide variety of virtual machines to run on this system. Examples
include:
❖ Eight virtual machines, each with a single virtual CPU and 4GB RAM
❖ Four virtual machines, each with a dual virtual CPU and 8GB RAM
❖ Four virtual machines, each with a dual virtual CPU and 16GB RAM
You might say, “Wait! Four times 16GB is 64GB, which is more than the 32GB of
RAM on the physical system. How is that possible?” The answer is that some virtualiza-
tion systems, such as VMware ESX, use advanced techniques (transparent page sharing,
ballooning, and swapping) to enable over-commit of memory. This is somewhat of a
mental somersault, because each virtual machine is probably managing memory inside
the operating system, and the virtual machine software is doing even more memory
management at the virtualization layer. This is all possible, although there might be
some performance issues when there are too many virtual machines sharing a limited
pool of physical memory.
Another misconception can be that hardware-assisted virtualization, such as that pro-
vided by the new class of CPUs, will make virtualization software obsolete, but CPU
virtualization is not the whole story. Device and I/O virtualization, as well as the rich set
of functionality provided by VMware Infrastructure, are critical to providing a complete
virtualization solution.
Over time, most people in the IT industry will come to understand these distinctions,
and you, as someone interested in deploying VMware Infrastructure, probably already
know the differences and capabilities, but having background information is always use-
ful.

Appendix B. Virtualization Ecosystem

A vast virtualization ecosystem has grown along with VMware, encouraged by published
VMware vSphere APIs to enable direct integration by manufacturers in security, man-
agement, storage, and networking. For example, leading security manufacturers such as
Symantec, McAfee, Check Point, Trend Micro, RSA, and Altor include VMsafe virtual
appliance plug-ins that automatically protect everything running within the vSphere
virtual machines without requiring the deployment of individual agents.
The three items included here are examples of these technologies. We have included spe-
cific examples from the host platform, networking, and software categories.

Hardware
Cisco is one of VMware’s key partners and produces two products specifically designed
to enhance a virtual infrastructure: the Nexus 1000V virtual switch and Unified Com-
puting System (UCS).

Cisco Nexus 1000V
The Nexus 1000V, also discussed in Chapter 11 on security, is a software switch built on
Cisco NX-OS specifically for VMware vSphere. It extends virtual networking to include
IP/port security rules, multi-host PVLAN, flow statistics, IGMP snooping, 802 1Q tag-
ging, rate limiting, and quality of service. The vSwitch and Port Groups are provisioned
along with the physical network infrastructure, using the familiar NX-OS CLI, enabling
more efficient configuration by the network team.

Figure 44: Cisco Nexus 1000V architecture

Virtualization Ecosystem / 105

The Nexus 1000V facilitates a higher percentage of server virtualization. DMZ ap-
plications, for example, can be virtualized with the help of private VLAN isolation and
security policy enforcement with ACL. Regulatory applications can be virtualized with
Netflow, ERSPAN, and port statistics that persist after vMotion, and Tier-1 applications
can be virtualized with increased visibility and I/O optimization with LACP, vPC host
mode.

Cisco Unified Computing System (UCS)

Figure 45: Cisco Unified Computing System B-Series (courtesy Brad Hedlund,
Cisco)
Cisco UCS is an example of an ecosystem computing platform architected from the
ground up to optimize the hosting of virtual infrastructure. It includes a hardware ver-

106 / Virtualization Ecosystem

sion of Cisco’s Virtual Network Link (VN-Link), enabling optimized I/O performance
while simplifying management with its tight integration with VMware vCenter Server.
Other innovations in extended memory capabilities, unified fabric support, and stateless
computing accelerate and simplify virtual infrastructure provisioning, performance, and
management. UCS is also the compute foundation for Acadia, the partnership between
VMware, Cisco, and EMC.

Software
There are a number of commercial and open source tools to manage a virtual infrastruc-
ture, including those that support a number of virtualization platforms. An example of
an open source tool is included here.

Cfengine
Cfengine was the first open source system configuration system, developed originally in
1993 as a research project by university researcher and professor Mark Burgess. It was
recently overhauled in 2008 as Cfengine 3 to take advantage of user experience and to
simplify its “promise language.” In 2009, a commercial version called Cfengine Nova
also became available, with enhancements and support.
Cfengine is a useful player in a virtualization strategy because it can configure systems
“hands-free” both on physical hosts and virtual guests, and its architecture can be fully
decentralized, allowing it to scale up to tens of thousands of machines, the kind of sce-
nario one expects to find in modern hosting centers and cloud computing dens. Cfen-
gine can recover from network outages and clock drift, so it can work well even when
systems are suspended for a length of time. Cfengine is used for deployments, change
control, and consistency or security checks, and it can also perform detailed performance
monitoring and compliance auditing.
Cfengine 3 has a few strengths that are useful for virtualization. It performs system con-
figuration and monitoring in a very lightweight fashion, with very little memory and
CPU overhead. This is especially useful in a virtualized environment where you want
to pack many virtual hosts onto limited hardware. Second, Cfengine’s now widespread
term convergence or computer immunology can bring a system from basically any initial
state into a state that is compliant with a defined policy. A long history has made Cfen-
gine available on all kinds of operating systems, including pretty much all UNIX vari-
ants, Linux, Macintosh, and Windows, and the latest commercial versions of Cfengine
Nova even run natively on Windows, supporting registry management and other Win-
dows oddities.
At the time of writing, there are plans to support VMware as well as other virtualization
engines directly in Cfengine, to simplify the deployment and management of virtual
machines. There are even ambitions to bring adaptive power-saving issues into the reper-
toire of ordinary mortals, as discussed in Green IT initiatives.

Manage Large Networks (MLN)
Manage Large Networks (MLN) is a management tool specialized for rapid deployment
and management of large numbers of virtual machines. It is vendor-agnostic, supporting

Virtualization Ecosystem / 107

VMware, Xen, and other virtualization technologies. Instead of using a graphical inter-
face, MLN utilizes a configuration language. This language allows accurate descriptions
of the virtual machines along with their internal configuration, such as users and net-
work configuration. A group of virtual machines is called a project, and the configura-
tion file describing the project can make use of super classes and inheritance, along with
variables, to shorten the amount of redundant text and remove errors. An extendable
plug-in framework allows users to extend the functionality of MLN.
MLN offers command-line management with the possibility of managing a project as an
atomic unit, e.g., starting, stopping, or polling status information on all virtual machines
belonging to the same project with one command. With the MLN daemon running,
projects can span several servers, allowing complicated scenarios to be realized effectively.
MLN was developed by Kyrre M. Begnum, PhD, associate professor at Oslo University
College, as part of his PhD program.

Glossary

In this glossary, multi-work terms that start with “VMware,” “vSphere,” or “vCenter” are
listed alphabetically using the word that follows in the term. “VMware vCenter term”
identifies technologies that are integrated with VMware vCenter, some of which are
considered add-on products requiring additional licensing. “VMware term” identifies
technologies that are stand-alone products. Where applicable, definitions are taken from
the formal VMware glossary guide.
A
VMware ACE—Virtualization product for enterprise desktop deployments providing a
highly configurable, secure, and portable PC environment.
ACE instances—The virtual machines that ACE administrators create, associate to vir-
tual rights management (VRM) policies, and package for deployment to users. An ACE
instance is an ACE.
VMware vCenter AppSpeed—An application performance monitoring tool engineered
specifically for multi-tiered applications. AppSpeed passively listens to traffic flowing
over a vSwitch (including the Nexus 1000V), which allows for discovery of transactions,
application mapping, performance monitoring against SLAs, and root-cause analysis.
Provides a method to evaluate performance of an application before and after virtualiza-
tion to ensure that performance remains consistent. This tool provides breadth in latency
analysis for an application.
C
VMware vCenter CapacityIQ—Identifies server resource inventories including used
and unused capacity. This can be used for capacity planning, budgeting, and lifecycle
management of resources. VMware vCenter Capacity IQ is used for cost avoidance and
justification, availability and risk mitigation, and project planning and decision-making.
VMware Capacity Planner—An agentless data collection and “what if ” scenario build-
ing tool that identifies server inventories and resource utilization to determine virtual
machine candidates, server consolidation ratios, and resource requirements for migrating
to a VMware Infrastructure based on target ESX host platform resources.
VMware vCenter Chargeback—Provides cost measurement, analysis, and reporting to
provide cost transparency and accountability for the virtual machines and the support-
ing virtual infrastructure. IT costs may be mapped to business units, cost centers, or
external customers to provide a better understanding of resource costs. This can further
be used to determine optimization for cost reduction.

110 / Glossary

VMware vCenter Chargeback API—Provides an interface for vCenter Chargeback
functionality. This includes management of the hierarchy, cost configurations, and re-
porting.
CIM Interfaces—Software interfaces designed for hardware management tool develop-
ment. This includes Server Management API—CIM SMASH interface to monitor and
manage virtualization server platforms and Storage Management API—CIM SIMI-S
interface to monitor and manage virtual storage.
Cluster—A server group in the virtual environment that enables a high-availability solu-
tion.
vCLI (vSphere Command Line Interface)—Allows you to manage your Virtual In-
frastructure using Windows PowerShell. This allows you to script and automate actions
you would normally do in vCenter. There are approximately 200 cmdlets (PowerShell
exposed procedures) to manage vSphere and ESX/ESXi functionality. There are many
pre-built scripts available online that can provide functionality such as finding all VM
snapshots, finding orphaned VMs, or even creating reports. Previously known as the VI
ToolKit.
VCB (VMware Consolidated Backup)—Provides the capability to perform SAN-based
backup and recovery of virtual machines using a backup proxy server without any net-
work or virtual machine overhead.
VMware Converter—Used for physical-to-virtual machine (P2V) migrations, as well
as importing virtual machines from other virtualization vendors (V2V). VMware Con-
verter can import multiple machines concurrently and non-disruptively. Designed for
large-scale consolidation, VMware Converter can be used with or without VMware
vCenter Server.
D
Datacenter—In the context of vCenter usage, an optional inventory grouping structure
contained within the datacenter structure. A vCenter Server supports multiple datacen-
ter folders. Datacenter folders can contain only datacenters and other datacenter folders.
vDR (VMware Data Recovery)—Provides data protection for virtual machines. VM-
ware Data Recovery is fully integrated with vCenter Server and includes data de-duplica-
tion to save on disk storage for full virtual machine backups. Includes file-level restore or
entire images as needed.
Data Source Name (DSN)—An ODBC (Open Database Connectivity) object that you
must configure to enable vCenter Server to access a database.
DPM (Distributed Power Management)—Dynamically starts up and shuts down ESX
host hardware to reduce power consumption.
DRS (Distributed Resource Scheduler)—Dynamically allocates and balances work-
loads across hosts in a cluster.
DV Port Group—A port group associated with a Distributed Virtual Switch (DVS). It
specifies port configuration options for each member port. It defines how connections
are made through the DVS to the network.

Glossary / 111

DV Uplink (DV Uplinks)—Physical uplinks attached to a vDS to enable VMs and
virtual network adapters connected to vNetwork Distributed Switch to connect to net-
works outside the hosts on which they reside.
DV Uplink Groups—Defines uplink policies for the associated DV Uplinks.
DV Port (Distributed Virtual Port)—A port on a DVS that connects to a host’s Ser-
vice Console or VMkernel or to a virtual machine’s network adapter.
DVS (Distributed Virtual Switch)—An abstract representation of multiple hosts defin-
ing the same vSwitch (same name and network policy) and port group. These represen-
tations are needed to explain the concept of a virtual machine being connected to the
same network as it migrates among multiple hosts.
E
Elastic Sky—The acronym ESX was created from the term Elastic Sky, a marketing
label that was created but not used by VMware. The initials plus the letter X became the
official product name. Elastic Sky was used as the name for a VMware band that was
founded by Jeff Hanson, John Arrasjid, Melinda Marks, David Haberman, Doug Clark,
Drew Kramer, and Ken Watson. Additional band members have participated with
Elastic Sky, including Tim Mann, Linda Pak, Norman Malonzo, Arisa Amano, Vittorio
Viarengo, and Robert Noth.
Emulation—In which a virtual machine simulates the hardware needed in a way that
allows it to run on a platform with a different CPU than it was originally designed to
work with.
VMware ESX/ESXi Server—ESX and ESXi are both hypervisors which install directly
on the server hardware. Although the deployment and management methods are slightly
different, both solutions provide better performance and availability than other meth-
ods. VMware ESX Server was initially released in 2002.
VMware ESX Server (VMware ESX Classic)—Classic ESX installs with a Linux-based
Service Console to assist with management functions.
VMware ESXi Server—A version of VMware ESX Server that may be installed like
VMware ESX Classic, on USB flash storage, or on an internal device. ESXi removes the
Service Console, reducing the attach surface due to a smaller footprint and allowing the
functionality to be embedded within the server hardware.
F
VMware Fault Tolerance or VMware FT—Provides clustering support of single vCPU
VMs without requiring the embedded application to be cluster aware. FT utilizes VM-
ware vLockstep technology. This technology uses an active secondary VM that runs in
virtual lockstep with the primary VM. VMware vLockstep establishes and maintains this
secondary VM. The secondary VM runs on a different host and executes the same set of
instructions, in the same sequence, as the primary VM.
Full Virtualization (Native Virtualization)—The guest OS is presented with a virtual
hardware abstraction that represents a physical machine. The virtual machine is recog-
nized and accessible to the operating system or applications software just as if it were a
physical machine, so no modification to the software is necessary. With full virtualiza-

112 / Glossary

tion, a standard operating system such as Windows or Linux, without modifications,
will run in a virtual machine.
VMware Fusion—A virtualization product for Intel-based Mac OS X systems.
G
GSX Server—Original name for VMware Server. GSX existed as a commercial product
from 2001 to 2006. The acronym is created from the term Ground Storm, a marketing
label that was created but not used by VMware. The initials were used and the letter X
was added to create the official product name.
GOS (Guest Operating System)—An operating system that runs within a virtual ma-
chine.
vSphere Guest SDK—Enables development of applications that will run within a vir-
tual machine using C or Java libraries. Enables customers to write smart applications
that respond to changes at the virtualization environment layer. Included with VMware
Tools.
VMware Guided Consolidation—Used for planning physical-to-virtual machine mi-
grations by utilizing VMware Capacity Planner Converter technology. VMware Guided
Consolidation is an optional vCenter component and is designed for small-scale consoli-
dation.
H
Hardware-assisted Virtualization—CPUs from Intel (Intel VT) and AMD (AMD-
V) implement hardware assistance for CPU virtualization; the first generation of these
CPUs was released in 2005 and 2006.
vCenter Heartbeat—Protects the vCenter Server, License Server and Database against
hardware, OS, application, and network downtime. Failover and failback are provided
for each. Protection is important especially when using VMware View, vCenter Lab
Manager, and vCenter SRM, which require vCenter to be running at all times.
VMware High Availability (VMware HA)—Provides automated restart of failed virtual
machines, regardless of the guest OS technology. Provides fault tolerance in the event of
an ESX host failure. VMware HA enables the automated restart of virtual machines on
other hosts in a cluster upon host failure, minimizing downtime without the cost of ap-
plication clustering.
Host—A compute platform supporting the execution of virtual machines. Includes
standard physical servers as well as platforms specifically designed to support virtual in-
frastructure such as Cisco UCS.
VMware Host Profiles—Enables the definition and application of standardized host
configurations. Also supports compliance checks against the defined standards.
Hosted Virtualization—Relies on having a standard operating system between the
physical computer and the virtualization layer. This requires installation of an operating
system such as Microsoft Windows or Red Hat Linux, and then installation of virtualiza-
tion software such as VMware Workstation on top of it. Finally, a guest operating system
such as Windows or Linux is installed in one or more virtual machines running within
VMware Workstation.

Glossary / 113

VMware Hyperic HQ—Provides complete discovery, monitoring, and analysis and
control of all application, system, and network assets both inside and outside the virtual
machines. Hyperic HQ includes full VMware ESX and VMware Server support, analysis
of utilization and performance within a VM, correlation of events between hosts and
guest OSes, and control of VMs. This tool provides detailed analysis of how the virtual
machine is performing. This tool provides depth in latency analysis for an application.
Hypervisor—Hypervisor virtualization platforms have a partitioning layer, which runs
directly on top of the hardware and below higher-level virtualization services, that pro-
vide a virtual machine abstraction. The hypervisor is installed on the computer, just as
though it is an operating system. It provides the capability to create virtual machine par-
titions, with a virtual machine monitor running within each partition.
The Hypervisors—A VMware band.
L
VMware vCenter Lab Manager—Provides a self-service portal for real-time provision-
ing, managing, and collaboration of virtualized development and testing environments.
VMware vCenter Lab Manager allows developers and testers to create and share libraries
of virtualized application environments used in software development and testing. Ap-
plications can be moved through lifecycle stages until they reach production state.
VMware vCenter Lab Manager SDK—Enables development of applications that use
Lab Manager Web service data, automate tasks, or integrate VMware Lab Manager with
software testing tools.
VMware Lifecycle Manager or VMware LCM—Manages the lifecycle of virtual
machines from request through provisioning and eventual archiving or destruction.
VMware vCenter Lifecycle Manager provides a self-service portal for virtual machine
requests, routed through a predefined workflow—streamlining provisioning, reducing
overhead, and providing consistent management of the virtual machine lifecycle.
Logical Partitioning (LPAR)—Found in mainframe computers such as IBM System z
(and, less commonly, on other IBM systems), as well as on computer systems from other
vendors. In logical partitioning the resources of a physical computer are partitioned so
that the computer’s memory may be split, allocating a specific range to each partition.
Hardware assistance is often used to partition a system but is not necessary for operating
system virtualization in general.
M
vSphere Management Assistant (vMA)—A Linux appliance with pre-built manage-
ment tools and the vCLI Interface. Allows scripting and agents to manage ESX, ESXi,
and vCenter Server systems. vMA is a virtual appliance that includes the vSphere SDK
and the vSphere CLI, logging capabilities, and an authentication mechanism.
Multipathing Policy—When connecting an ESX host to a Fibre Channel SAN, the
multipathing policy enables multipathing support to maintain a constant connection
between the ESX host and the storage device in the event that a critical connecting com-
ponent (e.g., host bust adapter, storage controller, storage processor, or Fibre Channel
cable) fails.

114 / Glossary

N
Network virtualization—VLANs (Virtual LANs) are used to segment networks on
physical networks. This is different from the virtualized network devices available in
VMware, although these two technologies can coexist.
O
VMware vCenter Orchestrator (vCO)—Provides out-of-the-box workflows to help
automate existing manual tasks. Workflows can be created, modified and extended to
meet custom needs.
VMware vCenter Orchestrator API—Allows for the programming of workflows for
execution by VMware vCenter Orchestrator.
P
Para-virtualization (OS-assisted virtualization)—The guest OS is presented with
a modified hardware abstraction. This requires operating systems to be modified and
ported to this particular virtualization platform. This reduces operating system com-
patibility, but that is a trade-off against potential increases in performance of certain
CPU-bound applications that run on systems without virtualization hardware. This
performance increase is achieved by hypercalls, a communication method that occurs be-
tween the guest OS and the hypervisor, but the performance advantage can vary greatly,
depending on the workload. However, each guest operating system, such as Linux, needs
to be modified. VMware has traditionally offered full virtualization, but aspects of para-
virtualization have been offered as an option for enhanced device drivers that increase
the efficiency of guest operating systems.
PC-over-IP (PCoIP)—PCoIP was a purpose-built protocol designed by Teradici to de-
liver a rich desktop experience consisting of content such as application windows, Web
pages, graphics, text, and streaming video and audio over both the LAN and the WAN.
PCoIP recognizes the different types of content and uses different compression algo-
rithms based on content type. PCoIP delivers multi-monitor support with up to 1920 x
1200 resolution, clear-type fonts, and 32-bit color per monitor for up to four monitors.
It includes Auto Display scaling, dynamic resizing, and support of monitor pivoting. It
enables multimedia redirection, USB support, support for host-based rendering of Flash,
and bi-directional audio.
VMware Player—Enables creating virtual machines and running virtual appliances.
Port Group—Specifies port configuration options, including VLAN tagging poli-
cies and bandwidth limitations for each member port. Network services connect to
vSwitches through port groups. Port groups define how a connection is made through
the vSwitch to the network. In typical use, one or more port groups is associated with a
single vSwitch.
Port Groups—A construct for configuring virtual network options such as bandwidth
limitations and VLAN tagging policies for each member port. Virtual networks that are
connected to the same port group share network policy configuration.
vSphere PowerCLI—Allows you to manage your Virtual Infrastructure using Windows
PowerShell. This allows you to script and automate actions you would normally do in
vCenter. There are approximately 200 cmdlets (PowerShell exposed procedures) to man-
age vSphere and ESX/ESXi functionality. There are many pre-built scripts available on-

Glossary / 115

line that can provide functionality such as finding all VM snapshots, finding orphaned
VMs, or even creating reports. Previously known as the VI ToolKit.
S
vCenter Server (VMware Server)—Free entry-level server virtualization product for
creating and running multiple virtual machines on existing physical Windows or Linux
servers. Formerly titled GSX Server.
Service Console—The command-line interface for an ESX server system that enables
administrators to configure the system. The Service Console is installed as the first com-
ponent and is used to bootstrap the ESX server installation and configuration. The Ser-
vice Console also boots the system and initiates execution of the virtualization layer and
resource manager. You can open the Service Console directly on an ESX server system.
If the ESX server system’s configuration allows Telnet or SSH connections, you can also
connect remotely to the Service Console.
VMware vCenter Site Recovery Manager (vCenter SRM)—Provides disaster recovery
workflow automation through a centralized management interface. SRM automates the
setup, testing, failover, and failback of virtual infrastructures between protected and re-
covery sites.
VMware vCenter Site Recovery Manager API—Provides an interface to SRM which
allows external management systems to initiate tests or failovers and record results.
Storage Virtualization—The process of abstracting a logical storage device from a
physical device. These are found in many areas, from virtual disks in VMware products
to Fibre Channel or IP network storage devices such as IBM’s SAN Volume Controller
(SVC), EMC Invista, or LeftHand Networks Virtual SAN Appliance (VSA). The physi-
cal location of the storage can be on a SAN, but the representation might be iSCSI. The
software handles the mapping between the physical storage and the logical storage.
Storage VMotion—Storage VMotion enables live migration of virtual machine disk
files across storage locations while maintaining service availability. Storage VMotion
utilizes VMotion technology to optionally move the VM to an alternate ESX host which
has access to both the source and target storage locations. Storage VMotion can move
the storage location of a virtual disk as long as the target is visible to the source and des-
tination ESX host(s). The processes of the corresponding VM can stay on the same host,
or the VM can be simultaneously VMotioned to a new host.
T
VMware ThinApp—Enables application virtualization, encapsulating the applications
from both the OS and each other. This eliminates conflicts from badly behaving applica-
tions as well as the requirement for regression testing.
V
vApp—Provides a logical entity, or object, comprising one or more virtual machines
using the OVF (Open Virtualization Format) to specify and encapsulate all components
of a multi-tier application. In addition, policies and SLAs can be associated with the ob-
ject as an attribute. The vApp construct is designed for interoperability of a multi-tiered
application on the virtual datacenter as well as for the ability to move the application
between internal or external clouds while maintaining the same SLAs.

116 / Glossary

Virtual Machine (VM)—This book focuses on “system virtual machines," a form of
virtualization whereby the underlying physical computer resources are mapped into one
or more different virtual machines (tightly isolated software containers that behave ex-
actly like a physical computer).
VMDirectPath—Offloads I/O processing from the hypervisor by allowing virtual ma-
chines to directly access the underlying hardware devices.
VMware Certified Professional or VCP—A certification designed for individuals wish-
ing to demonstrate expertise in virtual infrastructure and increase potential for career
advancement.
VMware Certified Design Expert or VCDX—The highest level of VMware certifica-
tion, evidencing an exceptional proficiency in designing and implementing successful
VMware infrastructures. John Arrasjid, lead author of this booklet, is VCDX #1 and
Duncan Epping is VCDX #7.
vCLI—The vSphere Command-Line Interface command set enables running common
system administration commands against ESX/ESXi systems from any machine with
network access to those systems. vSphere CLI commands are especially useful for ESXi
hosts that do not include a Service Console.
vCloud—The VMware vCloud initiative consists of technology from VMware as well
as the ecosystem of technology and cloud service providers to enable application delivery
on a common VMware vSphere platform. Nearly 1,000 validated applications are eas-
ily deployed to an on-premise environment or to a cloud without requiring recoding or
rebuilding.
vCloud API—Provides an interface for providing and consuming virtual resources
within a VMware-based cloud by enabling deployment and management of virtual-
ized workloads by working with vApps. This API is based on OVF standards providing
platform independence and multi-tenancy in a purely virtual infrastructure. Includes
functions for Inventory Listing, Catalog Management, Upload/Download/Provision-
ing Operations, vApp Configuration Operations, Resource Entities Operations, vApp
State Operations, and other operations. Also includes administrative functions including
Cloud, Org, vDC, Catalog, User, Group, and Role Administration.
vCloud Express—An Infrastructure as a Service (IaaS) offering providing pay-as-you-go
infrastructure that ensures compatibility with internal VMware environments.
vCPU—A virtual central processing unit is similar to the CPU in a traditional physical
machine. A virtual processor is assigned (either one or in multiples) to a virtual machine.
VMware View—A system for managing connectivity, security, and administration of
centralized virtual desktop computers hosted on ESX clusters. VMware View Manager
supports the connection brokering for the Virtual Desktop Infrastructure (VDI), while
View Composer provides advanced desktop image management.
VMware View Composer—Provides advanced desktop image management for the Vir-
tual Desktop Infrastructure (VDI).
VMware View Manager—Supports the connection brokering for the virtual desktop
infrastructure (VDI).

Glossary / 117

VIX—Allows development of programs and scripts to automate virtual machine and
guest OS operations. VIX runs on Windows or Linux platforms. It manages VMware
vSphere, ESX, ESXi, VMware Server, and VMware Workstation through the use of C,
Perl, and COM bindings, including Visual Basic, VBscript, and C#.
VIX API—Provides a programming interface to manage and automate functionality of
OS guests. Allows a script to be executing through vCenter using VMware Tools rather
than being sent across the VM networks.
Virtual Desktop Infrastructure (VDI)—Virtual Desktop Infrastructure is the term
coined by VMware to describe the process of running desktop operating systems and
applications as virtual machines on an ESX host.
Virtual Disk—A virtual disk simulates a physical disk in memory and is usable al-
though not physically present in the computer.
Virtual Disk Development Kit (VDDK)—Interface to allow ISVs to use VMDK as a
native format when developing virtual disk tools through the use of the VMware Virtual
Disk Libraries (VixDiskLib and ViMntapi).
Virtual Machine Monitor (VMM)—The VMM is a layer of software that runs be-
tween the hypervisor or host operating system and a virtual machine. It manages the
resources and their allocation to the virtual machines running on the system. The VMM
decouples the software from the hardware underneath. As computer scientist Butler
Lampson famously said, “All problems in computer science can be solved by another
level of indirection.” VMM’s decoupling capability provides substantial control over
how the guest operating system accesses the hardware.
Virtual Operating System—A small, lightweight component embedded with ThinApp
compiled applications.
Virtual SMP—Virtual SMP allows a single virtual machine to use multiple physical
processors simultaneously. It enables virtualization of even very resource-intensive appli-
cations such as databases, large Web servers, and ERP.
Virtual Switch—A software program emulating a physical switch to enable one virtual
machine to communicate with another. See vNetwork Standard Switch and vNetwork
Distributed Switch.
vLockstep—Virtual Lockstep technology is utilized as part of vSphere Fault Tolerance.
It facilitates zero downtime and zero data loss for a virtual machine by keeping a second
instance of the FT-enabled virtual machine in lockstep with the primary instance.
VMDK—The Virtual Machine Disk Format is a file format used to store a virtual ma-
chine image and the contents of a virtual machine hard drive.
VMFS—Virtual Machine File System is a cluster file system enabling storage of virtual
machine disk images, including snapshots.
VMkernel—The VMkernel is a microkernel handling CPU and memory directly as
part of CPU instructions. It provides a hardware simulation interface to guest systems.
VMmark—A benchmark tool specifically designed for measuring scalability of virtual-
ization host systems. Provides an accurate measurement of application performance in
virtualized environments. Measures virtual machine performance, determines how dif-

118 / Glossary

ferent hardware and virtualization platforms will affect performance, and enables “best
fit” choices for hardware. VMware is working with the Standard Performance Evaluation
Corporation (SPEC®) and members of the SPEC Virtualization subcommittee to de-
velop standard methods of comparing virtualization performance for virtualized applica-
tions running on hypervisors.
VMotion—VMware VMotion enables the live migration of running virtual machines
from one physical host to the other with zero downtime, continuous service availability,
and complete transaction integrity. VMotion migration requires either the same proces-
sor family on both the source and target ESX hosts, or “Enhanced VMotion Compatibil-
ity” (EVC) on a cluster of hosts with technologies enabling VMotion compatibility with
older servers. Hosts need to be grouped within the same vCenter datacenter. The shared
storage holding the VM virtual disk is presented to both the source and target hosts.
VMware VMsafe—Provides an open approach to security through an application
program interface (API). This enables selected partners to develop security products
for VMware environments. VMsafe gives fine-grained visibility over virtual machine
resources, making it possible to monitor every aspect of the execution of the system and
stop previously undetectable viruses, root kits, and malware before they can infect a
system. VMsafe provides inspection of virtual machine memory pages and CPU states,
filtering of network packets inside hypervisors as well as within the virtual machine itself,
and in-guest, in-process APIs that enable complete monitoring and control of process
execution. Guest virtual machine disk files can be mounted, manipulated, and modified
as they persist on storage devices.
VMsafe API—Allows vendors to develop advanced security products.
vNetwork—vNetwork refers to the collection of networking technologies enabling
optimal integration of networking and I/O functionality into vSphere. The vNetwork
enhancements include the vNetwork Distributed Switch, VMXNET3 (third-generation
para-virtualized NIC), IPv6 support extended to VMkernel and Service Console ports,
bi-directional traffic shaping, network VMotion, and VMDirectPath.
vNetwork API—Provides integration with the virtual networking capabilities of vSphere
to enable the development of advanced network tools.
vNetwork Distributed Switch (vDS)—Provides a switch that acts at a datacenter level
across multiple ESX hosts, providing centralized provisioning, administration, and mon-
itoring. Simplifies network management by moving the virtual network configuration
and management from the host level to the datacenter level.
vNetwork Standard Switch (vSS)—A software program emulating a physical switch
to enable one virtual machine to communicate with another. It is a basic Layer 2 switch
without routing.
vnicvNIC—A virtual network interface card is a virtual representation of a physical NIC
and is configured on top of a system’s physical network adapter.
VMware vShield Zones—Enforces corporate security policies at the application level in
a shared environment, while still maintaining trust and network segmentation of users
and sensitive data. Provides a mechanism to monitor, log, and block inter-VM traffic
with an ESX/ESXi host or between hosts in a cluster. This includes the ability to firewall,

Glossary / 119

bridge, or isolate virtual machines between multiple pre-defined zones. All activities,
blocked as well as allowed, are logged and can be graphed.
VMware vSphere—A collection of software providing management of a dynamic en-
vironment, cost reduction, and significant improvement to the life-work balance of IT
professionals. VMware vSphere is the next generation of the Virtual Infrastructure 3
product. Known also as the first cloud operating system, vSphere 4 was one of the most
complex software development projects of all time, consisting of over 3,000,000 engi-
neering hours by over 1,000 engineers over a three-year period.
vSphere SDK—A software development kit that acts as an interface ESX/ESXi, vCen-
ter, and VMware Server to extend the management of the Virtual Datacenter. Program-
ming languages supported include Perl, .NET, and Java.
vSphere SDK for Java—Supports simplified vSphere Management applications by
defining client-side data models. These models provide utility functions to simplify data
access to servers.
vStorage—Provides integration of advanced capabilities from storage vendors with the
vSphere Cloud OS from VMware. This API enables customers to leverage array-based
capabilities such as support for multipathing control, which enables advanced load bal-
ancing algorithms.
vStorage API—A collection of software providing management of a dynamic environ-
ment, cost reduction, and significant improvement to the life-work balance of IT profes-
sionals.
vswif#— The VMware vCenter Server Service Console NIC.
vSwitch—See Virtual Switch.
W
vSphere Web Services SDK—Provides a Web service accessible through the vSphere
API to develop client applications.
VMware Workstation—Enables running multiple virtual machines on a PC, each
with its own operating system. VMware Workstation resides on top of a host operating
system and supports over 200 guest operating systems. It includes clones, multiple snap-
shots, and secured “virtual rights management” features. See also VMware Player.

Booklets in the Series
#21: Cloud Computing with VMware vSphere, by John Arrasjid, Duncan Epping,
and Steve Kaplan
#20: Running the Numbers: System, Network, and Environment Monitoring, by About the Authors
Daniel V. Klein and John Sellens
#19: Configuration Management with Bcfg2, by Narayan Desai John Y. Arrasjid is a Principal Architect in the Technical Services
and Cory Lueninghoener Advanced Services group at VMware. He specializes in virtualization
aspects of business continuity, disaster recovery, and enterprise cloud.
#18: Deploying the VMware Infrastructure, by John Arrasjid, Karthik Balachandran, His experience includes work with AT&T, Amdahl, 3Dfx Interactive,
Daniel Conde, Gary Lamb, and Steve Kaplan Kubota Graphics, Roxio, and his own company, WebNexus Commu-
#17: LCFG: A Practical Tool for System Configuration, by Paul Anderson nications. As an architect, John develops IP for customers, field con-
sultants, and partners on topics including enterprise cloud, business continuity, disaster
#16: A System Engineer’s Guide to Host Configuration and Maintenance recovery, performance, security, and virtualized datacenters. John’s early contributions
Using Cfengine, by Mark Burgess and Æleen Frisch with the vmsnap/vmres scripts led to the development of VMware Consolidated Back-
#15: Internet Postmaster: Duties and Responsibilities, by Nick Christenson up. John regularly presents at VMworld, VMware Partner Exchange, USENIX Annual
and Brad Knowles Technical Conference and USENIX LISA, LinuxWorld, and other conferences. John is
also a founding member of the two VMware music bands Elastic Sky and The Hypervi-
#14: System Configuration, by Paul Anderson sors. John is a VMware Certified Professional and one of the first VMware Certified De-
#13: The Sysadmin’s Guide to Oracle, by Ben Rockwood sign Experts (VCDX 001). John holds a Bachelor of Science in Computer Science de-
gree from SUNY at Buffalo. He can be followed on Twitter at http://twitter.com/
#12: Building a Logging Infrastructure, by Abe Singer and Tina Bird vcdx001.
#11: Documentation Writing for System Administrators, by Mark C. Langston Duncan Epping is a Senior Consultant and the Datacenter Prac-
#10: Budgeting for SysAdmins, by Adam Moskowitz tice Lead for VMware. Duncan works primarily with enterprise
customers. He is focused on designing virtual infrastructures and
#9: Backups and Recovery, by W. Curtis Preston and Hal Skelly
specializes in business continuity, disaster recovery, and VMware HA.
#8: Job Descriptions for System Administrators, Revised and Expanded Edition, Duncan is a VMware Certified Professional and among the first VM-
edited by Tina Darmohray ware Certified Design Experts (VCDX 007). Duncan is the owner of
#7: System and Network Administration for Higher Reliability, by John Sellens Yellow-Bricks.com, one of the leading VMware/virtualization blogs worldwide and an
active contributor and moderator on the VMTN Community Forums. He can be fol-
#6: A System Administrator’s Guide to Auditing, by Geoff Halprin lowed on twitter at http://twitter.com/DuncanYB.
#5: Hiring System Administrators, by Gretchen Phillips Steve Kaplan is Vice President, Data Center Virtualization Prac-
#4: Educating and Training System Administrators: A Survey, by David Kuncicky tice for INX. Steve has authored scores of articles, white papers, and
and Bruce Alan Wynn books on different aspects of virtual infrastructure and is the author
of the VirtualMan comic book series. He has spoken on datacenter
#3: System Security: A Management Perspective, by David Oppenheimer, and desktop virtualization at venues around the globe and maintains
David Wagner, and Michele D. Crabb, and edited by Dan Geer Bythebell.com, a blog site emphasizing the economics of virtualiza-
#2: A Guide to Developing Computing Policy Documents, edited by tion. Steve holds a Bachelor of Science degree in business administration from the Uni-
Barbara L. Dijker versity of California, Berkeley, and an MBA from Northwestern University’s J.L. Kellogg
Graduate School of Management. He can be contacted at Steve.Kaplan@inxi.com or
followed on Twitter at http://twitter.com/roidude.

Use nix foundation for cloud computiing w. vsphere4

More Related Content

What's hot (20)

Similar to Use nix foundation for cloud computiing w. vsphere4 (20)

More from Venkata Ramana (7)

Recently uploaded (20)

Use nix foundation for cloud computiing w. vsphere4