Presentation architecting a cloud infrastructure

Architecting a
Cloud Infrastructure
Moderator:
Chris Colotti, VMware, Inc.
Aidan Dalgleish, VMware, Inc.
Duncan Epping, VMware, Inc.
David Hill, VMware, Inc.
Rawlinson Rivera, VMware, Inc.
INF-VSP1168
#vmworldinf

2
Disclaimer
 This session may contain product features that are
currently under development.
 This session/overview of the new technology represents
no commitment from VMware to deliver these features in
any generally available product.
 Features are subject to change, and must not be included in
contracts, purchase orders, or sales agreements of any kind.
 Technical feasibility and market demand will affect final delivery.
 Pricing and packaging for any new technologies or features
discussed or presented have not been determined.

3
Agenda
 Intros
 Gathering requirements
 Sizing and Scaling
 Host Design
 vCenter Design
 Cluster Design
 Networking and Security
 Storage
 Concluding

4
Introduction
 Chris Colotti
• Consulting Architect, VMware Global Cloud CoE
• VCDX 37 and owner of chriscolotti.us
• Twitter: @CColotti
 David Hill – (Not Available for VMworld US)
• Senior Solutions Architect, GTS
• Owner of Virtual-Blog.com
• Twitter @DaveHill99
 Aidan Dalgleish
• Consulting Architect, VMware Global Cloud CoE
• VCDX 10
• Twitter @AidersD

5
Introduction
 Duncan Epping
• Principal Architect, VMware Technical Marketing
• VCDX 7 and owner of Yellow-Bricks.com
• Twitter @DuncanYB
 Rawlinson Rivera
• Senior Consultant, VMware Professional Services
• VCDX 86 and owner of punchingclouds.com
• Twitter: @PunchingClouds

6
What is this session about?
 Architecting a Cloud Infrastructure
• See the CIM Case Study whitepaper. This is not a blueprint, it is an example!
 Design Decisions
 Real world examples
 Understanding the potential pitfalls
 Tweet / blog about this session, please use the #vmworldinf and
#VSP1168 hashtag
 Take pictures and tweet them!
• Best picture will get a free copy of the “vSphere 5.1 Clustering Deepdive”

8
Talk to your customer
 Most important part of any engagement
 Gather information and document!
 Categorize
• Requirements
• Nice to have
• Constraints
• Assumptions
 Conceptualize
 Sound like VCDX to Anyone?

9
Example Requirements
 Increasing agility / flexibility while decreasing cost of doing
business
 Availability of services defined as 99.9% during core business
hours
 Security compliancy requires network isolation for specific
workloads from other services
 Minimal workload deployment time
 Should be able to guarantee resources to groups of workloads as
part of internal SLAs
 Recovery time objective in the case of a datastore failure should be
less than 8 hours
 Servers hosted in the DMZ should be protected within the virtual
environment

10
Example Constraints
 Dell and AMD have been preselected as the platform of choice
 Eight 1GbE ports will be used per server
 NetApp’s NAS offering has been preselected as the solution of
choice
 All Tier 2 NAS volumes are de-duplicated
 Physical switches will not be configured for QoS
 Existing Cisco top of rack environment should be used for the
virtual infrastructure

11
Example Assumptions
 Sufficient Switch Ports are available
 Current storage infrastructure can handle expected workload
 Staff properly trained on vSphere
 …

12
Historical Best Practices
 It is best practice to have a 500GB LUN with a maximum of 15 VMs
 It is best practice to have an 8 host cluster
• vSphere
• vCloud Director (Fast Provisioning limitations)
 It is best practice to have a maximum of 4 VMs per core
 It is best practice to have a dedicated 1GbE link for vMotion
 It is best practice to have “high / medium / low” resource pools
 It is best practice to have an isolated management
 It is best practice to configure vMotion and Management on a VSS
and the other traffic on a VDS
 It is best practice to …

13
Some Use cases
 Server consolidation (power and cooling savings, green
computing, lowering TCO)
 OPEX Savings on redundant tasks
• Self Service Provisioning
 Server infrastructure resource optimization (load balancing, high
availability)
 Standardization
 Business Agility (Rapid provisioning)
 Infrastructure as a Service (IaaS)

14
Conceptualize Your Design
 Building blocks
• Operations
• Time to market
• Compliance

16
Basic details
 What does the environment look like today?
• How many sites?
• How many potential virtualization candidates?
• Multiple waves?
 How will this impact your Design / Project?
• Different Cluster / Datacenter structure
• Within the limits?
• Sizing based on X waves / years?
 What is the use case?
• Server consolidation?
• IaaS?
• Service Level Agreements (SLA)?

17
Tooling Options
 Use tools like
• VMware Capacity Planner
• PlateSpin Recon
• Lanamark
 Do we really need it?
• Don’t all results just look the same?
• What is important?
 What am I designing for?
• Average vs Peak
• Consolidation vs Performance

18
Compute Considerations
 How many eggs in on basket?
• Two sockets vs four sockets
• Optimal Memory configurations
• 8GB DIMMs are cheaper than 2 x 4GB
• Triple channel configurations
• Number of DIMM slots might be different per vendor / model
 AMD vs Intel
• AMD supports more cores, while Intel generally is faster
• VMmark can be used to make perf comparisons!
 TPS vs no TPS
• Using 64-bit Guest OS’es?
• Performance gain
 Sweetspot?
• Still seems to be dual socket – 96GB of memory

19
Network Sizing
 Is this ever really a bottleneck?
 In most of the Capacity Planner reports we’ve seen…
• Expected average network bandwidth requirement ~ 4Mbps based on an
average of 20 VMs per ESXi host.
 10GbE will lift all (or most) constraints for a very long time!
 Use the report to identify anomalies!

20
Storage Sizing
 Not only Size but Performance matters!
 (TOTAL IOps × % READ)+ ((TOTAL IOps × % WRITE) × RAID Penalty)
(42 x 62%) + (( 42 x 38%) x 2)
(26.04) + ((15.96) x 2)
26.04 + 31.92 = 57.96
 But what about size?
 How does this drive your Storage Considerations?

21
Storage Considerations
 RAID level used  impacts IOps
• IOps penalty can be severe
• Spindle count
 RTO  impacts amount of VMs per Datastore
• Backup environment need to be capable of restoring within RTO window
• (RTO * restore speed) / avg VM Size
• Don’t confuse Mb/S with MB/S!
 Adding it up…
• 270 VMs from backup perspective vs 50 VMs from IOps
• What does your customer feel comfortable with?
 Going SSD / Hybrid solutions?
• Potential undesired results – vCloud Director Catalogs

23
Design Considerations
 Vendor
 AMD vs Intel
 Blade vs Rack
• Density increases
• Hot spots
• Costs
• Management
 Additional considerations
• Is embedded ESXi available?
• How much local SSD (capacity and IOPS) can it handle?
• Does it have built-in 2x 10 GE ports?
• Does the built-in NIC card have hardware iSCSI capability?
• Management integration

24
ESXi boot considerations
 4 methods of booting ESXi
• Local Disk
• Local SD / USB
• SAN Boot
• PXE Boot with Auto-Deploy
• GUI Fling by Max Daneri!
 Considerations
• USB is cheap
• Local disk usually higher availability than USB
• SAN Boot easy to move identity, but what about costs
• Best of all worlds: PXE Boot!
• Brand new and dependencies…

25
Platform Composition: ESXi base,
drivers, CIM providers, …
Configuration: networking, storage,
date/time, firewall, admin password, …
What Is Auto Deploy
 No Boot Disk? Where does it go?
Boot Disk
All information on the state
of the host is stored off the
host in vCenter
Running State: VM Inventory,
HA state, License, DPM configuration
Event Recording: log files, core dump

26
What Is Auto Deploy
 No Boot Disk? Where does it go?
Boot Disk
All information on the state
of the host is stored off the
host in vCenter
vCenter ServerRunning State: VM Inventory,
Add-on ComponentsEvent Recording: log files, core dump
Host Profile
Image Profile

28
 How many VMs?
• Do I need a dedicated vCenter Server?
• Can I still use the vCenter Appliance?
 Is there a need for the Web Client?
• Can I use the vCenter Appliance for that?
• Use it!
 Will there be other products used like SRM / View /
vCloud Director?
• vCenter Heartbeat required?
 Statistic Levels will impact performance / scaling

29
Sizing vCenter and Update Manager
 Read the documentation!
• 50 hosts / 500 VMs  2vCPUs – 4GB
• 300 hosts / 3000 VMs  4vCPUs – 8GB
• 1000 hosts / 10000 VMs  8 vCPUs – 16GB
 Do we want to scale up or scale out?
 vSphere Update Manager on the same Server?
 How many users will be using vCenter?
 Use the Sizing Calculators for the database
• Consider Reservations

31
Where do we start?
 How many physical Datacenters will there be?
 Will each physical DC need a vCenter Server?
 For each vCenter, do we need multiple virtual Datacenters?
 For each DC, do we need multiple Clusters?
 For each Cluster, how many hosts?
Physical DC
vCenter
Datacenter Datacenter
Cluster Cluster
ESXi ESXi ESXi
vCenter
Datacenter
Physical
DC
vCenter Datacenter Cluster ESXi

32
 Separate clusters for DMZ?
• Why not use vShield App or vShield Edge?
 Separate clusters for test?
• Might also allow you to test vSphere patches!
 vCenter Datacenter object is a vMotion boundary not the Cluster!
 vCenter Datacenter object is a VDS boundary not the Cluster!
 Will you be using HA / FT / DRS / DPM?
• Did you know that each DRS cluster has its own thread on the vCenter server?
• Did you know that with vSphere 5 there’s a thing called Datastore
Heartbeating?
• There is no “primary / secondary” concept as of vSphere 5.0
• Admission Control is important!

33
 Is 8 the perfect Cluster size?
• Primary / secondary nodes (4.1 and prior) vs Master / Slave (5.x)
• Blade environment implication on design?
• LUN count vs Path count
• Linked Clones being used?
• DRS and DPM love big clusters
• HA benefits from big clusters
 What about EVC? Should I turn it on by default…
• EVC can only be enabled when all VMs are powered off
• Did you know that DRS requires EVC to be enabled to balance and place FT
virtual machines in a cluster?
 Is there a need for Resource Pools
• How will you handle shares?
• Reservations / limits?

35
 What type of vSwitch will be used?
• VSS vs VDS vs Cisco Nexus 1000v
• If vCloud Director what Network Pools are required?
 What are the pSwitch capabilities?
 Will VLANs be used?
 Will PVLANs be required?
• Consider vShield App?
 Requirements for Jumbo Frames?
• Helps support larger vCDNI packets and/or IP Storage
 What type of load balancing will be used?
• What type of traffic
• Load Based Teaming vs Virtual Port ID vs IP Hash
 Additional Security requirements?

37
 Network I/O Control
• Even in 1GbE environment NIOC is useful
• Especially when connecting outbound you want to
• Did you know that Limits apply to a NIC pair level?
• Did you know that shares apply on a NIC Port level?
 Additional Security Requirements?
• vShield App
• vShield App with Data Security
• vShield Edge
 3rd party security products?
• Hy-trust
• 2-factor authentication
• Audit trails
• Rigid, Hierarchical Access Controls

38
vShield App Considerations
 VNIC level firewall
 DVFilter used for in- and out-bound traffic
 vShield App Firewall per host!
 Set rules on vCenter Objects like Resource Pools and Portgroups
• Deploying a VM with the right rules is easy!
 Did you know that the vShield Manager cannot be locked out?
• The DVFilter is not applied to
the vShield Manager
 You can exclude VMs
from protection!

39
vShield Edge considerations
 Multiple edge security services in a single appliance
• Firewall (5 tuple)
• NAT
• DHCP
• VPN
• Load Balancing
 Network Isolation
 Data Security options
• Useful for compliance
 Think about “resiliency”!
• vSE HA from within vCD in 5.1
 vCloud Director leverages vShield
Edge heavily!
Tenant A Tenant C Tenant X
VMware
vShield Edge
VMware
vShield Edge
VMware
vShield Edge
VPNLoad balancerFirewall
Secure
Virtual
Appliance
Secure
Virtual
Appliance
Secure
Virtual
Appliance

40
vShield Manager considerations
 Local Database Backup
• vSM uses a MySQL local database
• If database is lost all configurations on vShield Edge are lost
 Availability considerations
• FT is supported (not from 5.1), and of course HA and VM Monitoring
• New 5.1 Appliance ships with 2 vCPUs
• vShield Manager Failure
• All existing, published, rules continue to be enforced
• All flow logging continues to be sent to syslog server
• No changes to rules or settings can be made
• Regular vShield Manager backups can be used to rebuild vShield Manager if needed
 Security considerations
• Default Passwords

42
 Protocol Wars!
 Multiple Tiers?
• Or even Auto-Tiering, what is the impact?
 vSphere Storage APIs – Array Integration (VAAI)
• Does it impact sizing?
 vSphere Storage APIs – Storage Awareness (VASA)
• Will it impact operations?
 Thin provisioning?
• Thin, Thick and Eager Zeroed Thick
• vSphere vs Storage Array!

43
 Can we use Storage DRS?
• Impact on storage array features?
• Impact on sizing?
• Impact on other VMware products like vCloud Director?
 Profile-Driven Storage?
• How does it utilize VASA?
 DR Requirements?
• Or possibly in the future?
 No more worrying about block sizes with VMFS-5
• When upgrading VMFS-3 to VMFS-5 block size does not change!
 Did you know VAAI is T-10 compliant?
• Makes leveraging it easier for lower-end devices

44
Impact of Features
 Storage DRS has constraints…
 SRM does not support Storage vMotion / Storage DRS
 vCloud Director does support Storage DRS in 5.1!
Feature or Product Initial Placement
Migration
Recommendations
Array-based replication
(SRDF, MirrorView, SnapMirror, etc )
Supported Manual: I/O and Space
Array-based snapshots Supported Manual: I/O and Space
Array-based Dedupe Supported Manual: I/O and Space
Array based thin provisioning Supported
Fully Automated: I/O and
Space
Array-based auto-tiering
(EMC FAST, Compellent Data Progression, etc)
Supported Manual: Space
Array-based I/O balancing
(Dell Equallogic)
Supported Manual: Space

FILL OUT
A SURVEY
EVERY COMPLETE SURVEY
IS ENTERED INTO
DRAWING FOR A
$25 VMWARE COMPANY
STORE GIFT CERTIFICATE

Architecting a
Cloud Infrastructure
Moderator:
Chris Colotti, VMware, Inc.
Aidan Dalgleish, VMware, Inc.
Duncan Epping, VMware, Inc.
David Hill, VMware, Inc.
Rawlinson Rivera, VMware, Inc.
INF-VSP1168
#vmworldinf and #VSP1168

Presentation architecting a cloud infrastructure

Recommended

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to Presentation architecting a cloud infrastructure (20)

More from xKinAnx (20)

Recently uploaded (20)

Presentation architecting a cloud infrastructure