Front cover

Enabling Hybrid Cloud Storage

for IBM Spectrum Scale Using
Transparent Cloud Tiering

Nikhil Khandelwal
Rob Basham
Sandeep R. Patil
Anbazhagan Mani
Amey Gokhale
Jinesh Shah
Kedar Karmarkar
Donald Mathisen
Larry Coyne
Arend Dittmer

In partnership with
IBM Academy of Technology

Redpaper
Notices

This information was developed for products and services offered in the US. This material might be available
from IBM in other languages. However, you may be required to own a copy of the product or product version in
that language in order to access it.

IBM may not offer the products, services, or features discussed in this document in other countries. Consult
your local IBM representative for information on the products and services currently available in your area. Any
reference to an IBM product, program, or service is not intended to state or imply that only that IBM product,
program, or service may be used. Any functionally equivalent product, program, or service that does not
infringe any IBM intellectual property right may be used instead. However, it is the user’s responsibility to
evaluate and verify the operation of any non-IBM product, program, or service.

IBM may have patents or pending patent applications covering subject matter described in this document. The
furnishing of this document does not grant you any license to these patents. You can send license inquiries, in
writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive, MD-NC119, Armonk, NY 10504-1785, US

INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES THIS PUBLICATION “AS IS”

WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED
TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A
PARTICULAR PURPOSE. Some jurisdictions do not allow disclaimer of express or implied warranties in
certain transactions, therefore, this statement may not apply to you.

This information could include technical inaccuracies or typographical errors. Changes are periodically made
to the information herein; these changes will be incorporated in new editions of the publication. IBM may make
improvements and/or changes in the product(s) and/or the program(s) described in this publication at any time
without notice.

Any references in this information to non-IBM websites are provided for convenience only and do not in any
manner serve as an endorsement of those websites. The materials at those websites are not part of the
materials for this IBM product and use of those websites is at your own risk.

IBM may use or distribute any of the information you provide in any way it believes appropriate without
incurring any obligation to you.

The performance data and client examples cited are presented for illustrative purposes only. Actual
performance results may vary depending on specific configurations and operating conditions.

Information concerning non-IBM products was obtained from the suppliers of those products, their published
announcements or other publicly available sources. IBM has not tested those products and cannot confirm the
accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the
capabilities of non-IBM products should be addressed to the suppliers of those products.

Statements regarding IBM’s future direction or intent are subject to change or withdrawal without notice, and
represent goals and objectives only.

This information contains examples of data and reports used in daily business operations. To illustrate them
as completely as possible, the examples include the names of individuals, companies, brands, and products.
All of these names are fictitious and any similarity to actual people or business enterprises is entirely
coincidental.

COPYRIGHT LICENSE:

This information contains sample application programs in source language, which illustrate programming
techniques on various operating platforms. You may copy, modify, and distribute these sample programs in
any form without payment to IBM, for the purposes of developing, using, marketing or distributing application
programs conforming to the application programming interface for the operating platform for which the sample
programs are written. These examples have not been thoroughly tested under all conditions. IBM, therefore,
cannot guarantee or imply reliability, serviceability, or function of these programs. The sample programs are
provided “AS IS”, without warranty of any kind. IBM shall not be liable for any damages arising out of your use
of the sample programs.

© Copyright IBM Corp. 2016, 2018. All rights reserved. 3

Trademarks
IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International Business Machines
Corporation, registered in many jurisdictions worldwide. Other product and service names might be
trademarks of IBM or other companies. A current list of IBM trademarks is available on the web at “Copyright
and trademark information” at http://www.ibm.com/legal/copytrade.shtml

The following terms are trademarks or registered trademarks of International Business Machines Corporation,
and might also be trademarks or registered trademarks in other countries.
AIX® IBM Spectrum™ Redbooks®
Global Technology Services® IBM Spectrum Accelerate™ Redpapers™
GPFS™ IBM Spectrum Archive™ Redbooks (logo) ®
IBM® IBM Spectrum Control™ Storwize®
IBM Cloud™ IBM Spectrum Scale™ System z®
IBM Elastic Storage™ POWER8® XIV®

The following terms are trademarks of other companies:

Linux is a trademark of Linus Torvalds in the United States, other countries, or both.

Microsoft, Windows, and the Windows logo are trademarks of Microsoft Corporation in the United States,
other countries, or both.

Java, and all Java-based trademarks and logos are trademarks or registered trademarks of Oracle and/or its
affiliates.

Other company, product, or service names may be trademarks or service marks of others.

4 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Enabling Hybrid Cloud Storage for IBM Spectrum Scale
Using Transparent Cloud Tiering
This IBM® Redbooks® publication provides information to help you with the sizing,
configuration, and monitoring of hybrid cloud solutions using the transparent cloud tiering
(TCT) functionality of IBM Spectrum™ Scale. IBM Spectrum Scale™ is a scalable data, file,
and object management solution that provides a global namespace for large data sets and
several enterprise features.

The IBM Spectrum Scale feature called transparent cloud tiering allows cloud object storage
providers, such as IBM Cloud™ Object Storage, IBM Cloud, and Amazon S3, to be used as a
storage tier for IBM Spectrum Scale. Transparent cloud tiering can help cut storage capital
and operating costs by moving data that does not require local performance to an on-premise
or off-premise cloud object storage provider.

Transparent cloud tiering reduces the complexity of cloud object storage by making data
transfers transparent to the user or application. This capability can help you adapt to a hybrid
cloud deployment model where active data remains directly accessible to your applications
and inactive data is placed in the correct cloud (private or public) automatically through IBM
Spectrum Scale policies.

This publication is intended for IT architects, IT administrators, storage administrators, and

those wanting to learn more about sizing, configuration, and monitoring of hybrid cloud
solutions using IBM Spectrum Scale and transparent cloud tiering.

Introduction
Transparent cloud tiering is a new Cloud Services feature available with version 4.2.1 of IBM
Spectrum Scale that lets you integrate object storage as a storage tier, leveraging the full
life-cycle management capabilities of the IBM Spectrum Scale Integrated Lifecycle
Management (ILM) policy engine to control the movement of data. This publication covers
features available with version 5.0.0 of IBM Spectrum Scale. Transparent cloud tiering works
with many object storage providers, and is a near-line storage tier useful as a target for
inactive data.

© Copyright IBM Corp. 2016, 2018. All rights reserved. ibm.com/redbooks 1

 Here are the guidelines on how to read this paper, which is organized into the following
sections:
򐂰 Technology Overview: Target Audience: Everybody.
򐂰 IBM Spectrum Scale transparent cloud tiering use cases: Scenarios where cloud tiering
would typically be used. Target Audience: Pre-sales, solution architects.
򐂰 Sizing Guidelines and Deployment Models: Guidance on how to plan for the
implementation of transparent cloud tiering and recommendations on how to deploy the
service. Target Audience: Solution architects, administrators.
򐂰 Configuration Best Practices: Recommended settings and tunable parameters when using
the transparent cloud tiering. Target Audience: Solution architects, administrators.
򐂰 Setting Up cloud tiering: How to set up policies for movement of data and routine
administration of the tiering service covering such things as periodic clean-up of deleted
and re-versioned files. Target Audience: Administrators.
򐂰 Monitoring: How to monitor the tiering service. Target Audience: Administrators.
򐂰 Security: Security considerations for the tiering service. Target Audience: Administrators.

Technology overview
This section provides an overview of IBM Spectrum Scale, cloud object storage, and how IBM
Spectrum Scale provides a cloud storage tier by using the fully integrated transparent cloud
tiering feature.

According to International Data Corporation (IDC), the total amount of digital information
created and replicated surpassed 4.4 zettabytes (4,400 exabytes) in 2013. The size of the
digital universe is more than doubling every two years and is expected to grow to almost 44
zettabytes in 2020.

Although individuals generate most of this data, IDC estimates that enterprises are
responsible for 85 percent of the information in the digital universe at some point in its
lifecycle. That means organizations take on the responsibility for designing, delivering, and
maintaining information technology systems and data storage systems to meet the demand.

Both IBM Spectrum Scale and object storage, including IBM Cloud Object Storage, are
designed to deal with this growth. The transparent cloud tiering function combines the two.
IBM Spectrum Scale can take advantage of object storage such as IBM Cloud Object Storage
as a storage tier using transparent cloud tiering service.

In a cloud deployment, a hybrid cloud model blends elements of both the private and the
public cloud. In the simplest terms, the hybrid model is a private cloud that allows an
organization to tap into a public cloud when and where it makes sense. IBM Spectrum Scale
transparent cloud tiering supports tiering of data to public cloud object store like IBM Cloud
and Amazon S3, allowing you to adopt a hybrid cloud strategy where data is placed in the
correct cloud model.

IBM Spectrum Scale overview

IBM Spectrum Scale is a proven, scalable, high-performance file management solution. It
provides world-class storage management with extreme scalability, flash accelerated
performance, and automatic policy-based storage that has tiers of flash through disk to tape.

2 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 IBM Spectrum Scale reduces storage costs up to 90% while improving security and
management efficiency in cloud, big data, and analytics environments.

IBM Spectrum Scale provides software-defined storage that can manage quintillions of files
and yottabytes of data. It provides high performance, simultaneous access to all of your data
in a single global namespace.

IBM Spectrum Scale software has been designed to work across multiple platforms,
supporting a mixture of IBM AIX®, Linux, Linux for IBM System z®, and Microsoft Windows
clients with support for flash and spinning disk storage. Integrated protocol access allows
NFS, SMB, Object, HDFS, and native POSIX clients to seamlessly access a shared global
namespace.

With transparent cloud tiering, Spectrum Scale now can seamlessly and transparently tier
data to cloud object store like IBM Cloud, Amazon S3, and even to native object stores like
IBM Cloud Object Storage.

Figure 1 shows an overview of IBM Spectrum Scale.

Client workstations Users and applications

Compute Farm

Single namespace
POSIX SMB/CIFS OpenStack
Map Reduce
Connector Cinder Swift
NFS Manila Glance

Site B
Site A

Spectrum Scale
Automated data placement and data migration

Site C

Tape IBM Cloud Object Storage

Flash
Storage Rich Amazon S3
Servers
Disk Cloud

Figure 1 Spectrum Scale Overview

Note: IBM Spectrum Scale offers a no-cost try and buy IBM Spectrum Scale Trial VM. The
Trial VM offers a fully preconfigured IBM Spectrum Scale instance in a virtual machine,
based on IBM Spectrum Scale GA version. You can download it by clicking the Start your
free trial button.

IBM Spectrum Scale deployment models

IBM Spectrum Scale as software-defined storage provides various deployment models to suit
your workload needs. The following deployment models are described in this section:
1. IBM Elastic Storage™ Server (ESS): An IBM specialized hardware solution based on IBM
Power with IBM Spectrum Scale declustered RAID technology. The ESS can be deployed
as a spinning-disk or a high-density all-flash solution.

3
 2. Shared SAN storage: The cluster is based on traditional shared SAN storage, and can
include disk and flash storage systems.
3. Storage-rich servers: The cluster is based on storage rich servers that take advantage of
the Spectrum Scale file placement optimizer feature.

IBM Elastic Storage Server

ESS is software-defined storage that combines IBM Spectrum Scale (formerly IBM GPFS™)
software and associated clustered file systems with the CPU and I/O capability of the IBM
POWER8® architecture.

Elastic Storage Server is a high performance, highly available, and scalable IBM Spectrum
Scale building block, meeting today’s needs for high performance and business analytics
applications. Replacing hardware-based disk RAID controllers, the IBM Spectrum Scale
declustered RAID delivers superior data protection and rebuild times that take a fraction of
the time that is needed by hardware-based RAID controllers.

ESS is ideal for large data sets due to its high-density storage options, and superior
scalability. Combined with transparent cloud tiering, this deployment model is preferred when
you want to minimize the frequency of recalls and data transferred to and from the cloud to
save on network and cloud provider costs. IBM ESS comes in two models, GSxS and GLxS,
where the key difference is the type and capacity of disk, making it suitable for different
workloads. For more information, see the specifications of IBM ESS.

Figure 2 depicts the deployment of the IBM Spectrum Scale cluster over IBM ESS.

SpectrumScale
Protocol/Client Nodes

InfiniBand / 40 GigE / 10 GigE

ESS
…

Spectrum Scale Cluster

Figure 2 IBM Spectrum Scale with IBM ESS

4 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
All Flash ESS
The GSxS models of ESS with high-density SSDs can deliver high IOPS performance for big
data, analytics, HPC and other workloads that require high IO rates with low latency. Using
high capacity 4 or 15 TB SSDs, the solution provides a large amount of storage capacity in a
compact footprint. IBM Spectrum Scale native RAID functionality provides industry-leading
reliability, availability, and serviceability (RAS) capabilities and fault-tolerance.

In combination with transparent cloud tiering, the all-flash storage can be configured with a
deep cloud storage tier to store cold data in a two-tier solution, or combined with other ESS
models to provide a multi-tier solution. All-flash ESS is recommended when you have a large
number of files, low application latency, and smaller total data sizes that need to be
maintained on the primary storage.

Note: An IBM Spectrum Scale Failure Group is defined by the administrator and contains
one or more Network Shared Disks (NSDs). Failure groups are used for replication of data
and metadata for reliability.

Shared storage
In this deployment model, a few NSD servers have direct access to shared storage system
through a SAN or shared connection, and IBM Spectrum Scale client nodes access the
storage through the NSD servers, as shown in Figure 3. This model is typically used with
backend storage providing RAID functionality.

S S S C C C

FC Switch

RAID Storage

pectrum Scale Cluster

S Server C Client
Figure 3 IBM Spectrum Scale with SAN-based shared storage

5
Storage rich servers
IBM Spectrum Scale File Placement Optimizer (FPO) capability is a share-nothing-cluster
deployment model that uses networked storage rich servers, as shown in Figure 4. IBM
Spectrum Scale FPO is suitable for workloads that are read dominant or benefit from data
locality such as Hadoop. Here the data protection is achieved by using replication of data
across the cluster.

Spectrum Scale Cluster

10 GigE/1 GigE

Storage-Rich Servers
S S S F - IBM Power
F
- X86

10 GigE / InfiniBand

S Server License F FPO License

Figure 4 IBM Spectrum Scale with storage rich servers

Cloud storage overview

Object storage is the primary method for accessing data used in the cloud, and it is
increasingly used for on-premises solutions. Object storage is growing for these reasons:
򐂰 It is designed for large-scale deployments in many ways (multi-site, multi-tenant, massive
amounts of data).
򐂰 It is easy to use and yet meets the growing demands of enterprises across a broad
spectrum of applications and workloads.
򐂰 It allows users to balance storage cost, location, and compliance control requirements
across data sets and essential applications.

Object storage has a simple REST-based API. Hardware costs are low because it is typically
built on commodity hardware. One of its key limitations is that most of the object storage
services are only eventually consistent. In order to accommodate its massive scale of data
and widespread geographic locations, object storage service might not immediately reflect all
updates. The lag is typically small, but can be noticeable during network failures.

Object storage is typically offered as a service on the public cloud like IBM Cloud or Amazon
S3, but is also available as on-premises systems like IBM Cloud Object Storage. For more
information about object storage, see the IBM Cloud Object Storage Overview.

Due to its characteristics, object storage is becoming the predominant storage for active
archival of unstructured data both for public and private clouds.

6 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 Note: IBM Spectrum Scale also supports object storage as one of its protocols. One of the
key differences between IBM Spectrum Scale Object and other object stores like IBM
Cloud Object Storage is that IBM Spectrum Scale Object comes with a unified file and
object access with Hadoop capabilities. It is also suitable for high performance-oriented or
data lake use cases, whereas IBM Cloud Object Storage is more of a traditional cloud
object store suitable for the Active Archive use case.

Cloud deployment models

Cloud object storage can be deployed both on and off premise.

Off-premise cloud
Off-premise public clouds, such as IBM Cloud Object Storage or Amazon S3, provide storage
options with minimal extra equipment and datacenter costs. The ability to rapidly provision,
expand, and reduce capacity, along with a pay-as-you-go model, make off-premise clouds a
flexible option. When considering off-premise clouds, it is important to consider all the costs
and the pricing models of the cloud solution.

Many cloud providers charge a monthly fee for the amount of data stored in the cloud. In
addition, many clouds charge a fee for data that is transferred out of the cloud. This charge
model makes public clouds ideal for data that is infrequently accessed. Storing data that is
accessed frequently in the cloud can result in higher costs. It might also be necessary to have
a high-speed dedicated network connection to the cloud provider.

On-premise cloud
On-premise cloud solutions, such as IBM Cloud Object Storage, provide flexible, easy-to-use
storage with attractive pricing. On-premise storage is ideal for cases where you need to have
the data under your own control or where you have higher recall rates from on-premises
applications that would make off-premise cloud storage expensive.

For a cloud solution with multiple access points and IP addresses, a load balancer is required
for high availability and throughput. The transparent cloud tiering service provides a simple
load balancing service. Also, several commercial and open source load balancers are
available. Contact your cloud solution provider for supported load balancers.

IBM Spectrum Scale transparent cloud tiering: What it is and

how it works
Transparent cloud tiering is a fully integrated service of IBM Spectrum Scale that provides a
native cloud storage tier. IBM Spectrum Scale administrators can use ILM policy semantics to
move files into and out of cloud storage. When migrated to cloud storage, the file name
remains in the file system global namespace and the data can be retrieved transparently
because the files are accessed by standard applications. Data can be moved between local
storage and object storage by policy, on access, or by using command line tools.

7
 Figure 5 shows an overview of how the transparent cloud tiering service is provided.

IBM
POSIX NFS CIFS OBJECT Public
Cloud

IBM Spectrum Scale Cloud Amazon

Cluster Services S3

IBM
COS
Tape
Fast Slow
Flash Private
Disk Disk
Cloud

Figure 5 Capabilities with transparent cloud tiering

The transparent cloud tiering service runs on groups of IBM Spectrum Scale nodes and
provides the following characteristics:
򐂰 Scalability: Scales the service up by defining for the cluster the number of node groups
(each of which can have a number of nodes) that are needed to achieve the performance
target.
򐂰 Efficient: Data moves directly from IBM Spectrum Scale to cloud storage without passing
through an intermediate application or appliance. This configuration means less processor
burden for the transfer and for tracking what has moved.
򐂰 Transparent Recalls: If a file is accessed that has been migrated to the cloud, it can be
automatically recalled.
򐂰 Highly available: No single point of failure and automatic failover when a node goes down.
򐂰 Low administrative overhead: Policy-based movement means that after the system is
configured, there is no need for constant intervention. Lifecycle management of deleted
and reversioned files can also be automated.
򐂰 Secure: All data at rest and on the wire is encrypted. Both a local key manager and IBM
Security Key Lifecycle Manager (ISKLM) key management support are available.
򐂰 Data Integrity: All data transfers are checked by using object storage integrity checking
techniques.

Note: Protocol nodes are also referred to as Cluster Export Services (CES) nodes in
IBM Spectrum Scale documentation.

8 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Transparent cloud tiering major components and terminology
This section describes the IBM Spectrum Scale transparent cloud tiering components. A
typical multi-node configuration is shown in Figure 6 with terminology referenced throughout
the document.

Transparent cloud tiering major components and terminology

Cloud Object Storage Cloud Object Storage

Provider 1 Provider 2

Cloud Account Cloud Account Cloud Account

Container Container Container Container
Pair Pair Pair Pair

Cloud Storage Access Point Cloud Storage Access Point

Cloud Service Cloud Service

Node Group Node Group
Node1 Node 2 Node3 Node 4 Node 5 Node 6 Node 7 Node 8 Node 9

Cloud Cloud Cloud Cloud Cloud Cloud

Cloud Service Cloud Cloud
Service Service Service Service Service
Client Node Client Client
Node Node Node Node Node

NSD Storage nodes IBM Spectrum Scale

Cluster
File System File System File Set
CES Protocol nodes

Figure 6 IBM Spectrum Scale cluster with Cloud Service nodes

Note: The transparent cloud tiering service runs on one or more Cloud Service Node
Groups that consists of Cloud Service Nodes. For resiliency and availability reasons, a
group typically contains more than one node. In Figure 6, the file systems, the Cloud
Service Node Groups, and the Cloud Accounts, and associated transparent cloud tiering
containers are all colored to show their groupings (connected by the arrows).

The objects in Figure 6 have these definitions:

Cloud Service Node Group
The transparent cloud tiering service runs on the Cloud Service Node
Group. The node group allows for high-availability in failure scenarios
and balancing the workload across all nodes in the node group, which
in turn allows for faster data transfers. A node group can only be
associated with one IBM Spectrum Scale file system, one Cloud
Account, and one data container on the Cloud Account.

9
 Cloud Service Node A Cloud Service Node is where the actual work is done. Requests for
transfer go to these nodes and they perform the transfer. A Cloud
Service Node can only be defined on an IBM Spectrum Scale Protocol
Node.
Cloud Storage Access Point
One or more URLs or IP addresses used to access a cloud object
storage provider.
Cloud Account Object Storage is multi-tenant but transparent cloud tiering can only
communicate with the tenant on the object storage that is represented
by the Cloud Account.
Cloud Client The remainder of the nodes in the cluster can act as cloud clients
(provided the operating system is supported). This light-weight client
can be used to send cloud tiering requests to the transparent cloud
tiering service. Each Cloud Service Node comes with a built-in Cloud
Client. A cluster can have as many clients as needed. Remotely
mounted clients are also supported for transparent recall operations.
Migration Movement of file data from the IBM Spectrum Scale cluster to the
cloud object storage.
Recall Movement of file data from cloud object storage to the IBM Spectrum
Scale cluster.

IBM Spectrum Scale transparent cloud tiering use cases

This section showcases use cases that can be broadly applied to a number of workloads:
򐂰 Active Archive
򐂰 Analytics

Before you go into the use cases, consider these key questions that can help determine
whether a workload is right for transparent cloud tiering:
򐂰 Do you require long-term retention of data?
򐂰 Does your data consist of larger size files or objects, such as unstructured objects,
images, movies, or documents?
򐂰 If you plan to deploy your own object storage for use with transparent cloud tiering, do you
have access to sufficient object storage capacity?
򐂰 Do you have requirements for data security, availability, scalability, and cost-effectiveness?
򐂰 Do you have plans to procure the infrastructure required to support cloud storage,
including network connectivity and load balancing services, and private or public cloud
access?

Active Archive
The growing amount of unstructured data in the data centers, and the need to manage and
store the data efficiently is what is driving the need for Active Archive. Active Archive is a
method of tiered storage that gives a user access to data with an online or near-online
capability.

10 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
With an Active Archive, applications can directly recall archived data based on metadata
stored in IBM Spectrum Scale, maintaining a global namespace even for data that has been
moved to the cloud. An Active Archive can facilitate more frequent and unpredictable data
access when compared to a pure Archive solution.

An Active Archive solution using IBM Spectrum Scale transparent cloud tiering has many
advantages over other archive products:
򐂰 Flexible deployment options (private cloud, hybrid, or public cloud).
򐂰 Cloud solutions such as IBM Cloud Object Storage are highly available across multiple
sites with geo-dispersed erasure coding. Many public cloud providers also provide high
availability with multiple site protection. Such storage pairs nicely with IBM Spectrum
Scale synchronous replication, which allows for support across multiple sites to allow
uninterrupted access to the cloud data tier even if an entire site goes down.
򐂰 Intermittent bulk migrations and recalls can result in faster data access times than a pure
archive solution.
򐂰 Reduced capital equipment costs with off-premise cloud solutions.

Active Archive can benefit various industries as shown in Figure 7.

Need for Active Archive across multiple industries

Sample Use Cases

Financial Manufacturing

• Document Repository • Industrial Images

• Audio recordings (Quality monitoring) • CCTV Video captures
• CCTV Video captures • Document repository
• Transaction Records (Compliance archival) • IoT Data

Healthcare Media and

Entertainment
• DNA Sequence Data Storage • Image / Photos
• Biological Datasets Storage • Video content
• Genome Data Access • Audio Storage
• Medical Imaging • Game data

Figure 7 Active Archive use cases

As already outlined, IBM Spectrum Scale with transparent cloud tiering can be a good match
to service these Active Archive use cases. See Figure 8 on page 12.

By contrast, IBM Spectrum Archive™ is a more standard archive service that is

recommended for standard archive use cases:
򐂰 Store infrequently used data on tape
򐂰 Lowest total cost of ownership (TCO)
򐂰 Long-term retention
򐂰 Tape loading can cause longer access times

Note: Transparent cloud tiering cannot be used to manage the same file system that is
being managed by IBM Spectrum Archive.

11
 Ideal for active archive uses cases and
transparent tiered storage

Global namespace

CIO Finance Engineering

transparent
IBM Spectrum Scale cloud
tiering

Flash Disk
Gold Pool Silver Pool

Tier 1 Tier 2

Figure 8 Transparent cloud tiering with IBM Cloud Object Storage

Using transparent cloud tiering along with IBM Cloud Object Storage achieves these goals:
򐂰 Storage Efficiency (primary storage can be freed up for high-performance applications)
򐂰 Cost reduction (infrequently accessed data can be moved to a lower-cost tier, resulting in
cost savings)
򐂰 Data Protection (multi-site protection and access for archive data)
򐂰 On-Demand Access (ready for faster recalls and unpredictable data access patterns)
򐂰 Metadata search
򐂰 Limited user intervention (policy-based management)

Active Archive for digital media libraries

A digital media library is a huge repository of unstructured data, predominantly in form of
images and videos. The data undergoes various processes throughout its lifecycle, from
content capturing, to editing, to automated digital manipulation, and finally to publication and
delivery of the content. This process creates a huge number of by-products and different
versions of content in the form of images, videos, and project files that are prime candidates
for archiving.

Such workloads demand high-performance storage and archival of the large amount of data
to keep down costs. IBM Spectrum Scale with transparent cloud tiering is an ideal storage
system for such workloads. IBM Spectrum Scale provides interfaces (including NFS, SMB,
Object (S3/Swift), and POSIX) that can be used for data ingestion and publication in a global
namespace.

12 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
IBM Spectrum Scale client nodes provide performance unmatched by other protocols, which
is ideal for image and video processing.

With transparent cloud tiering, an administrator can easily move processed data, its input,
and its by-products to an object store archive (such as IBM Cloud Object Storage) or to public
cloud storage (such as IBM Cloud or Amazon S3). An administrator can use the power of IBM
Spectrum Scale policies to automatically manage the data to be moved to the object storage
provider. Active archiving frees up the primary Spectrum Scale storage so that it is always
ready for incoming media.

In Figure 9, the objects (images and videos) are ingested by using IBM Spectrum Scale
Object or file interfaces. Because IBM Spectrum Scale provides a single global namespace,
the ingested media objects are immediately available to media processing compute farms
running the IBM Spectrum Scale client. The final product, which resides on IBM Spectrum
Scale, is then made available to the consumer through one of the Spectrum Scale interfaces
like object or file (NFS, SMB).

After the project has been delivered, it can be moved through transparent cloud tiering to a
cost efficient, durable Active Archive such as IBM Cloud Object Storage, IBM Cloud, or
Amazon S3. This solution can be deployed at a single site as shown in this diagram, or across
multiple sites using Spectrum Scale synchronous mirroring and IBM Cloud Object Storage
erasure coding. This configuration ensures that uninterrupted access to data continues during
a total site failure.

Figure 9 shows the Active Archive media use case.

metacafe

YouTube Dailymotion
Media House

Publishing Channels
VM Farm
Ingest for video processing whereevery VM
Media Objects has a Spectrum Scale client
Final Video
available for streaming
Virtual Virtual
Object Object NFS/SMB Machine …. Machine
Instances Instances

Spectrum Scale

Media Content available for processing

Content Ready for Archive via Spectrum Scale clients

Final content sent for publishing

IBM Cloud Object Storage (Cleversafe)

or
Cloud storage (Amazon S3)

Figure 9 Active Archive for Media

13
Analytics
With features such as high-speed clustered data access, HDFS transparency, and native
POSIX access, IBM Spectrum Scale is an ideal storage platform for analytics workloads. It
can deliver results fast with in-place data analytics and can provide a single data repository
for content creation, as well as servicing analytics applications such as Hadoop and Spark.

Using IBM Spectrum Scale, you can create a custom architecture to match your analytics
workload. You can use storage rich servers, ESS, and flash mixing storage technologies as
needed. Transparent cloud tiering can be used to extend these capabilities. In this example,
IBM ESS GSxS is the primary storage building block, providing low-latency high-density flash
storage for analytics. In addition, IBM Cloud Object Storage provides a repository for cold
data, which transparent cloud tiering can recall if it is required for further analysis at any time.

The solution, as shown in Figure 10, scales to multiple petabytes of data, with the active data
on high-speed flash. This high-speed, low-latency flash solution provides an ideal analytics
platform to deliver fast results. Inactive data is automatically moved to object storage, which
frees up space on the high-performance flash and delivers significant cost savings. The
tiering is handled by IBM Spectrum Scale policies, so the administrative effort needed to
manage the data placement is low.

High Performance Analytics

Users and Applications Faster insights

from data to Users and Applications
the users

Analytics (Hadoop/Spark)
Hadoop Object NFS/SMB Hadoop Object NFS/SMB
Recall Data
CES (TCT) Node CES (TCT) Node
NSD Node NSD Node
IBM Cloud Object
File System X Storage
IBM Spectrum Scale Migrate Data

Analytics (Hadoop/Spark) natively supported on IBM

ESS ESS Spectrum Scale provides faster insights Into the data

Figure 10 Analytics use case with transparent cloud tiering

Unsupported use cases of transparent cloud tiering

Transparent cloud tiering is not recommended for a certain set of use cases:
򐂰 Using transparent cloud tiering to migrate/recall hot (active) data.
The cloud storage tier should be used as a repository for infrequently accessed data.
Migrating frequently accessed data causes multiple recalls, which can result in
unacceptable latency and unwanted cloud provider data recall access costs.
򐂰 Using transparent cloud tiering as a backup mechanism.
Transparent cloud tiering is designed to place data in less expensive storage. It is not a
replacement for a backup solution. With proper planning, transparent cloud tiering can be
used to restore previous versions of a file or deleted files from a cloud tier. See “Restoring
files from the cloud tier” on page 22 for additional information.

14 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 򐂰 Using transparent cloud tiering for disaster recovery scenarios.
Metadata on the IBM Spectrum Scale file system is required for transparent cloud tiering.
Simply placing the data in the cloud is not a complete DR solution.
򐂰 Using transparent cloud tiering when the average file size is smaller than the
recommended minimum average file size, which is currently 1 megabyte.
When a file is migrated, some metadata remains on the IBM Spectrum Scale file system.
Space savings for small files are not as great with transparent cloud tiering. In addition, the
additional processor cycles required to create and move small objects can result in lower
throughput.

Sizing Guidelines and Deployment models

This section is designed for architects and administrators to help with sizing and deploying a
Spectrum Scale solution with transparent cloud tiering. It outlines what can affect the load of a
cloud tiering solution and how it can affect hardware requirements.

On any transparent cloud tiering deployment, it is important to consider these items:

򐂰 The number of cloud service nodes and node hardware
򐂰 The network connection between the IBM Spectrum Scale cluster and the cloud storage
provider
򐂰 The active storage and metadata requirements for the IBM Spectrum Scale cluster
򐂰 The capabilities of the cloud object storage that is used in the deployment

Things to consider
The following pieces of information can help you determine the hardware and software
needed to successfully run transparent cloud tiering.

Amount of data to migrate

Consider the amount of data that will be moved to and from the cloud. Analyze the existing
size of the file system or data set that you are considering when tiering to the cloud, and how
the data is used.

Number of files in file system or file set

Examine the size and number of files that are in the file system. The number of files affects
how long it will take to execute a policy to migrate data to the cloud. Also, consider future
plans for file system growth.

File size distribution

Consider the size of the files that will be migrated to the cloud. After a file has been migrated,
a small amount of metadata will remain on the file system. The minimum metadata size is the
size of the file system inode (4096 bytes by default). As a result, small files experience less
capacity savings than larger files. In addition, larger files make better use of network
bandwidth, although very large files can also lead to excessive recall times.

15
 Recall frequency
Consider how often data that is migrated to and from the cloud provider is likely to be
accessed. Most public cloud providers charge for data as it is copied from the cloud. Data that
is migrated to cloud storage should be accessed infrequently. For existing IBM Spectrum
Scale file systems, the policy engine can be used to generate lists of files that have not been
accessed recently. For policy examples, see the IBM Spectrum Scale wiki on the IBM
developerWorks site.

Application tolerance for latency

If an application accesses a file that has been migrated to the cloud, it can be recalled
transparently. However, because the file is located in the cloud, additional latency occurs as a
result of this recall. Larger files and slower network speeds result in higher recall latency
because the transfer of files from the cloud storage to the IBM Spectrum Scale cluster takes
longer to complete.

Transparent cloud tiering checks the integrity of each file as it is recalled. The integrity check
requires the entire file, so the tiering service does not allow an application to access a file until
the file has been recalled in its entirety. Table 1 provides a sample of latency versus file size
for a reasonably fast cloud service configuration.

Note: This example is intended to get you thinking about latency. Your actual performance
might be better or not as good as the numbers listed here, depending on your specific
environment.

Table 1 makes the following assumptions:

򐂰 Overall recall performance is file size dependent, with the best performance occurring at
larger file sizes. To keep the example simple, a fixed number of 700 MBps is used.
򐂰 For this example, assume that 32 files are being recalled on a node, dividing the overall
throughput between all 32 parallel transfers.

Note: The presence of intermittent heavy recall workloads where recall requests get
queued up can lead to longer recall times.

Table 1 is an example of how latency increases directly and linearly with file size.

Table 1 Latency related to file size

File Size Latency

2 MB 0.1 seconds

20 MB 1 second

200 MB 10 seconds

2 GB 100 seconds

20 GB 17 minutes

200 GB 170 minutes

16 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 Encryption/key manager
For environments that encrypt data that is sent to the cloud, transparent cloud tiering can use
either a local key or an external key manager, such as IBM Security Key Lifecycle Manager
(ISKLM). If you use a local key manager, ensure that the key is backed up properly to allow
access to encrypted data during a node failure.

Resource requirements and considerations

This section describes the hardware and other infrastructure that are required for transparent
cloud tiering.

Cloud service nodes

Cloud services run on CES nodes, and must meet the minimum requirements for CES
protocol nodes on IBM Spectrum Scale. See the IBM Spectrum Scale FAQ for current CES
protocol node requirements and other hardware restrictions for cloud tiering. IBM Spectrum
Scale 4.2.2.4 supports up to 4 file systems, each managed by 1 - 4 cloud service nodes.

For a redundant solution with high availability, a minimum of two cloud service nodes per file
system are required. For on-premises clouds, or cloud service providers with high-bandwidth
networks such as multiple 10 GbE networks, four cloud service nodes might improve
throughput from a file system to the cloud provider.

If the cloud service node is used for only cloud services, the minimum memory requirements
of a protocol (CES) node are sufficient. However, if the node is used for other protocols (for
example, serving NFS or SMB exports) or for running other applications, ensure that the
system has sufficient CPU and memory capacity for a node servicing multiple protocols.

In most cases, the cloud service nodes start scans of the file system to determine which files
to migrate to the cloud. As a file system contains more files, this scan becomes more I/O
intensive due to the increased metadata, and might have to be run less frequently. If
implementing cloud tiering on a file system with hundreds of millions or billions of files, extra
memory devoted to the scan might improve scan times.

Transparent cloud tiering attempts to use as much bandwidth as possible when transferring
files to and from the cloud. As a result, it might impact other operations that use the same
network adapter. Avoid sharing a network that IBM Spectrum Scale is using for internal
communication with cloud tiering. It is also best to avoid sharing the network adapter used by
protocol services running on the cloud service node with cloud tiering.

Network
Network requirements and bandwidth vary significantly for onsite versus off-site cloud
providers. Data volumes and application performance requirements also play a major role
when sizing the network used for transparent cloud tiering.

Ensure that the network bandwidth is capable of handling the amount of data that will be
migrated to the cloud. You can calculate the projected network requirements by examining the
data that is generated and changed in your environment over time. Using this approximation,
divide the amount of data by the expected network bandwidth to ensure that the data can be
migrated in the expected time.

17
 For example, if your applications generate 10 TB of data to transfer to the cloud per day and
the data link between the IBM Spectrum Scale cluster and your cloud provider is 1 Gbps
(approximately 100 MBps), it will take approximately 10,000,000 MB / 100 MBps =
100,000 seconds to migrate the data. With only 86,400 seconds in a day, it will take more time
to migrate the data than is available, so a faster network is required.

Recall latency times are equally important to consider, especially for large files. For example,
if you recall a 60 GB file over the same 1 Gb network link, it will take at least 60,000 /
100 MBps = 600 seconds to recall the file. This time can increase if multiple files are being
read at the same time, or if the network has other traffic. In many cases, long recall times are
acceptable because they occur infrequently. Always consider the needs of your applications
and users.

When working with an offsite cloud, work with your cloud provider to determine the bandwidth
that can be obtained to the cloud. In most cases, this amount depends on a WAN connection
to an external network. Some providers, such as IBM Cloud and Amazon, might offer
dedicated network links to provide more predictable, higher speed access.

Onsite cloud storage systems, such as IBM Cloud Object Storage, can have multiple
redundant access points, each with a different IP address. Because transparent cloud tiering
can currently only be configured with a single IP address, a load balancer is required for
redundancy that can improve system performance by using multiple object endpoints. Several
software and hardware load balancers are available. Contact your cloud storage provider for a
list of recommended options.

IBM Spectrum Scale backend storage

Transparent cloud tiering works on a standard IBM Spectrum Scale file system so it can be
used with any backend storage. Cloud tiering uses IBM Spectrum Scale policy scans to scan
file system metadata and determine what files might be candidates to move to cloud storage.
As the number of files in the file system grows, the size of the metadata increases and the
time that is required to run a policy scan increases.

High-performing metadata storage can help reduce scan times and the impact of scans on
other system operations. With larger deployments of this service, using flash storage for
metadata can greatly reduce the time and impact of these scans.

Cloud storage
Transparent cloud tiering supports the Amazon S3 and OpenStack Swift protocols for object
storage. It has been tested with IBM Cloud Object Storage and IBM Spectrum Scale Object,
as well as with IBM Cloud Object and Amazon S3 public clouds. Depending on the cloud
solution, it might be necessary to work with your solution provider to ensure that the solution
offers the capacity and throughput required by the file workload, particularly for on-premises
or dedicated cloud solutions. For on-premises cloud deployments, a load balancer might be
required if there are multiple endpoints to meet redundancy and performance requirements.

When connecting to public clouds, generally one Cloud Storage Access Point connection is
sufficient because public storage always comes with built-in load balancing. For private
on-premise object storage, it is advisable to have more than one Cloud Storage Access Point
connection which will allow the transparent cloud tiering service to provide basic load
balancing. An alternative would be to get an on-premise load balancer though in our
experience the cost vs. the benefit trade-off doesn't pay off and the built-in load balancer will
be the best option.

18 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Configuration guidelines
This section provides the administrator with guidelines for deploying the cloud tiering solution.
It includes recommended tunable parameters in different components of IBM Spectrum Scale
and transparent cloud tiering.

File system guidelines

Cloud tiering uses native IBM Spectrum Scale APIs to read and write data from the file
system. Therefore, it is important to ensure that the file system read/write performance is
good. Setting the following attributes with proper values is key in achieving optimal
performance for tiering.

All of these attributes can be set by using the mmchconfig IBM Spectrum Scale command.
The following settings can improve the performance of cloud service nodes:

Note: If the cloud service node is also being used for other workloads, such as protocol
exports or application I/O, then the configuration needs to take into account those
workloads as well.

pagepool This setting is the IBM Spectrum Scale memory that is used for caching file
data and file system metadata. Set an appropriate value based on the IO
pattern (Random, Sequential, Random Direct, NSD servers).
Tiering many smaller files (Less than 1 MB) to cloud storage results in a
random IO pattern, whereas tiering larger files (1 MB or greater) to cloud
storage results in a sequential IO pattern. For random IOs, this setting
should be set to higher values depending upon the overall memory
available on the cloud service node.
dmapiWorkerThreads
Cloud tiering accesses data from the IBM Spectrum Scale file system
through portions of the data management API (DMAPI) layer. This setting
defines the performance of this layer. The default value is 12. Set this
attribute to a higher value, such as 60, to improve migrate and recall
performance.
workerThreads This setting is used to tune the file system performance in environments
that are capable of high sequential or random read/write workloads, or
small file activity. It automatically sets the values of 20 other parameters. As
of Spectrum Scale 4.2.1, it is recommended to not change any other
thread-related parameters if the value of this attribute is changed to a
non-default value.

For details about these attributes, see the IBM Spectrum Scale wiki.

CES preferred practices

Cloud services can be run on IBM Spectrum Scale CES nodes. For environments with high
bandwidth requirements to the cloud provider, it might be beneficial to use a non-CES node,
or to ensure that protocol services, such as NFS, SMB, or Object on the CES nodes, run on
different network adapters than cloud services. If you are running cloud services with the
protocol services on the same node, that node might require extra CPU and memory
resources to avoid performance degradation.

19
Network preferred practices
The two network communication paths that are used by transparent cloud tiering are shown in
Figure 11. Cloud service and cloud client nodes communicate through the cluster-wide
network to coordinate the migration and recall of files from a cloud service provider.

Transparent cloud tiering traffic along this network is minimal, and places little extra demand
on the cluster-wide IBM Spectrum Scale network. Cloud service nodes are required to
communicate with the cloud provider to perform migrations and recalls of file data. The cloud
network can generate significant traffic when migrations or recalls are in progress.

Cloud Object Storage

Cloud Network
Node 1 Node 2 Node 3 Node 4 Node 5 Node 6 Node 7 Node 8 Node 9

Cloud Service Node Group

Cloud Cloud Cloud Cloud Cloud

Cloud Cloud Cloud Cloud
Service Service Service Service
Client Client Client Client Client
Node Node Node Node

Cluster Network

Figure 11 Transparent cloud tiering network

Although both the cloud network and cluster network can coexist on the same adapters, it is
best to have a separate adapter running on a different subnet than cluster traffic to
communicate to the cloud object storage. If protocol services, such as NFS, Object, or SMB,
are running on the cloud service nodes, keep protocol communication and cloud
communication on separate adapters as well to avoid performance degradation due to
network contention.

On many networks, it is common to bond multiple network adapters on a single bonded

adapter for higher performance and availability. Transparent cloud tiering supports such
bonding. For availability, active/passive bonds are sufficient for connectivity during a network
port or cable failure. For performance, it is common to use a bonding mode such as 802.3ad.
If you use a network bond, pay particular attention to the hash mode on the network adapter
and on the switch.

In most cases, Transparent cloud tiering communicates with a single endpoint. Therefore,
hashing modes that rely on a MAC address or an IP address might not distribute network
traffic across multiple adapters evenly. Refer to your network switch or Operating System
vendors for details about hashing modes.

20 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Policy execution guidelines
Policies can be used to migrate files from local storage to the external cloud storage pool. The
following section describes guidelines when working with IBM Spectrum Scale cloud tiering
policies. For sample policies, see “Setting up cloud tiering” on page 24. It also describes the
two ways in which policies can be applied: Manual policy execution or threshold-based
execution.

Policy parameters
To use multiple cloud service nodes efficiently, the following settings can be applied when
running any policy using the mmapplypolicy command:
򐂰 -m is the number of threads that are used within each mmapplypolicy process during the
policy execution phase. The default value is 24. It is recommended to set this value to a
number equal to the number of available CPU cores to speed file system scan times.
򐂰 -B is the number of files that are passed for each invocation of the transparent cloud tiering
migration command. The default value is 100. Based on the average file size, this number
can be changed so that each node processes an equal and reasonably large number of
files in a single iteration.

For example, when migrating 200 files through four cloud service nodes, the default value of
-B (100) tries to distribute 100 files to each node. In this case, only two nodes are used to
transfer files. Decreasing this value to 50 forces each node to receive 50 files, which
equalizes the load across all of the nodes.

Running policies on multiple nodes

When running migration or recall policies, it is important to consider which nodes are used to
execute the policies. By default, the policy is executed on the node where the mmapplypolicy
command is run. More than one node can be passed to the mmapplypolicy command with the
-N flag to improve scan performance. This section describes how to choose those nodes.

Data migration is performed by all of the nodes that participate in executing the policy. So if a
cloud gateway node runs the policy, that node performs the migration of the file to the cloud
provider. If a non-cloud gateway node participates in the migration, a file migration request is
forwarded to a random cloud gateway node for processing. Therefore, it might optimize
bandwidth and improve balance to just use the cloud gateway nodes in the node group that
manages the file system to execute the policy.

Invoking mmapplypolicy
The mmapplypolicy command can be invoked to either run a policy that has been established
for a file system using mmchpolicy, which is usually the case in threshold migrations, or can
be run with a policy file to perform migrations. Using multiple nodes to run a policy can
improve the speed of the policy run. For guidance on which nodes to use, see the “Running
policies on multiple nodes” on page 21. When running a policy using more than one node, a
global work directory is required in order for the nodes to coordinate work.

Typically, this can be a temporary directory on the file system being scanned, for example:
/gpfs/gpfs0/.mmapplypolicy

To invoke mmapplypolicy using cloud service nodes, a typical command would be:
mmapplypolicy gpfs0 -g /ibm/gpfs0/.mmapplypolicy -N TCTNodeclass

21
 Where TCTNodeclass is the configured node class for the cloud service nodes, and the file
system in this case is named gpfs0. This invocation would run the established policy, but a file
containing an arbitrary policy can be passed by using the -P flag.

Threshold migration
If threshold migrations are being used, it is useful to automatically run mmapplypolicy when a
pool crosses a threshold. To do so, a callback must be added that invokes mmapplypolicy
during low space and out of space events. See the threshold migration section for more
details about enabling these events. The callback for low space and out of space events
should be configured to run mmapplypolicy on all TCT nodes. This configuration can be done
by using the callback parameters. For example:
mmaddcallback migration --event lowDiskSpace,noDiskSpace -N TCTNodeclass --command
/usr/lpp/mmfs/bin/mmapplypolicy --parms “%fsname -N TCTNodeclass -g
/ibm/gpfs0/.mmapplypolicy --single-instance”

The invocation for this configuration contains the TCTNodeclass twice. The first command
specifies which nodes the callback is run on. In this example, all of the nodes in the
TCTNodeclass class are notified of the event to ensure that the policy is run even if only a
single node is unavailable. The second command specifies which nodes will be used to run
the policy itself. In this example, the nodes in the TCTNodeclass class are again used to run
the policy. Also, note the --single-instance option, which ensures that only one instance of
the migration runs at a time.

Restoring files from the cloud tier

While transparent cloud tiering is not a replacement for a backup solution, it can be used to
restore previous versions of files or files that were accidentally deleted from a IBM Spectrum
Scale file system.

Transparent cloud tiering maintains a Cloud Directory database, which tracks files in the IBM
Spectrum Scale file system that have been migrated to the cloud tier. Transparent cloud
tiering does not delete files in the cloud tier that have been deleted on the IBM Spectrum
Scale file system or remove older versions of files stored in the cloud until a reconcile process
is run. As a result, older versions of files can be restored from the cloud.

File versions available in the cloud tier can be listed with the following command:
mmcloudgateway files cloudList --path <path> --file-versions <filename>

If a file is lost, or an older version of a file is required that exists in the cloud tier, it can be
recovered using the mmcloudgateway files restore command.

If a significant number of files or an entire directory tree are removed and need to be restored,
it might be necessary to recreate the directory structures as well as restoring multiple files in
parallel. For guidance and assistance with this, see “Bulk file restore” on page 28 for further
guidance and examples.

IBM Spectrum Scale transparent cloud tiering guidelines

Transparent cloud tiering has several configurable options. In most cases, the default options
work well. If needed, the following tunable parameters can be changed for transparent cloud
tiering.

22 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 These parameters are configured by using the mmcloudgateway config set command. For
more information about this command, see IBM Knowledge Center:
migrate-threadpool-size
This parameter directly corresponds to the number of CPU cores on the cloud
service node, which processes the incoming migrate requests. Generally, set this
value equal to the number of CPU cores on a cloud service node. The default value
is 32. Having a smaller value queues up the incoming data migration requests.
recall-threadpool-size
This parameter directly corresponds to the number of CPU cores on the cloud
service node, which processes the incoming recall requests. Generally, set this
value equal to the number of CPU cores on a cloud service node. The default value
is 32. Having smaller value queues up the incoming data recall requests.

IBM Cloud Object Storage transparent cloud tiering guidelines

IBM Cloud Object Storage (COS) is an ideal private or managed cloud object solution for use
with transparent cloud tiering. COS is an ideal target for inactive data. If you use COS as the
cloud object provider with transparent cloud tiering, it has these considerations:
򐂰 Synchronize the time on the COS Accessors and the cloud service nodes by using
Network Time Protocol (NTP) servers. If the Accessor and cloud service nodes are not in
sync, credential errors might occur when authenticating with the accessor.
򐂰 Transparent cloud tiering creates a vault on COS using the COS default vault template. A
specific vault template can be specified in the mmcloudgateway account create command.
򐂰 The COS user requires the Vault Provisioner role, which is assigned through the dsNet
Manager UI.
򐂰 Make sure that an access pool is created and available on the COS system. It must have
Cloud Object Storage as the API type.

The recommended vault settings for COS and TCT are shown in Table 2.

Table 2 Recommended vault setting for COS and TCT

Configuration Recommended Values Comment

Width As per IBM Cloud Object Storage

documented configuration for production

Threshold As per IBM Cloud Object Storage

documented configuration for production

Write Threshold As per IBM Cloud Object Storage

documented configuration for production

Alert Level As per IBM Cloud Object Storage

documented configuration for production

SecureSlice Technology Disabled Transparent cloud tiering performs the

client-side encryption of all data before the data
is stored on the object storage. Therefore,
encryption is disabled on the vault.

SecureSlice Algorithm Not applicable because SecureSlice is

disabled

23
 Configuration Recommended Values Comment

Versioning Disabled Transparent cloud tiering has built-in versioning

capability, so IBM Cloud Object Storage
versioning can be disabled.

DeleteRestricted Yes/No Gateway does not attempt to delete the vaults,

so this setting can be set to Yes or No.

Name Index Disabled Disabling this setting can result in improved

Vault performance.

Recovery Listing Enabled When Name Index is disabled, this setting is

enabled to be able to conduct a disaster
recovery if needed.

For the latest information about COS with transparent cloud tiering, see IBM Cloud Obect
Storage considerations.

Setting up cloud tiering

This section describes how to configure IBM Spectrum Scale with transparent cloud tiering.

Cloud tiering policies

The IBM Spectrum Scale policy engine is a key part of file management for transparent cloud
tiering. Policies can be used for batch file migration and recall operations. A new policy
feature introduced with transparent cloud tiering is a transparent recall policy that is applied
for automatic data recall on file access.

Batch migrate policies

IBM Spectrum Scale offers a rich policy language to help identify files to migrate. Many
policies have identify files based on parameters, such as last access time, file heat, fileset, file
path, and file extension. For a full list of attributes available, see the policy section of IBM
Knowledge Center for IBM Spectrum Scale.

A policy is executed by using the mmapplypolicy command. There are three ways to start a
policy scan:
򐂰 Based on a capacity threshold
򐂰 On a schedule, for example daily or weekly
򐂰 Manually by a system administrator by running the mmapplypolicy command

Transparent cloud tiering includes several sample policies in the /opt/ibm/MCStore/samples

directory. These policies include both threshold and directory-based migration. These are
only samples. You can create custom policies with whatever rules and attributes meet your
needs.

Before running any migration policy for the first time or installing a new rule on a file system
with mmchpolicy, run mmapplypolicy with the -I test flag. When the test flag is used, the scan
phase of a policy is run, but no files are migrated. This process allows you to see how much
data would have been migrated and a list of files that will be migrated.

24 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 Use this data to determine whether the policy is performing as expected, or if changes are
required before applying the policy. Testing a policy can protect against accidental file
migrations and recalls.

Threshold-based migrations
A threshold-based migration typically specifies a file system storage pool, a high-water mark,
and a low-water mark, with an additional optional premigration threshold. If the storage pool
usage exceeds the high-water mark, the policy attempts to migrate files until the pool usage is
at the low-water mark. The optional premigrate threshold attempts to migrate more files to the
cloud provider, and leaves a copy of the file in the local file system. Premigrates can be used
to speed future migrations by transferring file data while still leaving the file in active storage.

When using a threshold policy, a migration is complete when the low water mark is reached.
Because this amount is usually not all of the files that could be migrated, use a weight to
determine which files go first. The weighting parameter (weight()) can be based on many file
attributes, including file heat, file size (example: move largest files first) or last access time
(example: move oldest files first). Example 1 is a sample threshold migration policy that is
weighted on last access time (ACCESS_TIME).

Example 1 Sample threshold migration policy

/* Define a cloud pool, using the 'mmcloudgateway' command to perform migrations to */
/* the pool */
RULE EXTERNAL POOL 'cloudpool' EXEC '/usr/lpp/mmfs/bin/mmcloudgateway files' OPTS '-F'

/* Define a migrate rule, in this example we will migrate from the system pool */
RULE 'MigrateToCloud' MIGRATE FROM POOL 'system'
/* Define the threshold to use, in this example we will migrate when the pool is */
/* 85% full, and attempt to migrate files until the pool is 75% full */
THRESHOLD (85, 75)

/* Next, define a weight for the file to determine which files to migrate first */
/* Here we will use the last access time for the file by subtracting it from the */
/* current time, giving a higher weight to older files */
WEIGHT(CURRENT_TIMESTAMP - ACCESS_TIME)

/* Define that we want to move to the cloud pool */

TO POOL ('cloudpool')
/* We don't want to migrate files that are smaller than 8k in size, because there */
/* are little to no space savings due to file overhead */
WHERE (KB_ALLOCATED > 8)
/* Do not migrate cloud services internal files, which reside in the */
/* .mcstore directory */
AND NOT(PATH_NAME LIKE '%/.mcstore/%'

The rule in Example 1 migrates all of the files in a file system, from the least recently
accessed to the most recently accessed, until the system pool is less than the 75% threshold.
Threshold-based migrations can be run manually, on a schedule (by using cron for example),
or in response to an event such as low disk space. To configure a threshold policy to
automatically run when a file system is low on disk space using a callback, complete the
following steps:
1. Ensure that the enableLowspaceEvents configuration parameter is set to yes in
mmlsconfig. If not, run the mmchconfig enableLowspaceEvents=yes command.

Note: The IBM Spectrum Scale cluster must be restarted for this change to take effect.

25
 2. Create a migration callback. See the migration callback guidance in “Running policies on
multiple nodes” on page 21.
3. Create a file that contains the threshold policy, the recall policies for the file system, and
other policies required for the file system (such as a placement policy). Apply this policy to
the file system by using the mmchpolicy command or the Spectrum Scale GUI.

Note: It is always recommended to test a new policy by using mmapplypolicy -I test

before applying it to a file system.

Remember that when you apply a threshold policy to a full file system for the first time, the
initial data movement might take a long time. For example, if a 100 TB file system is 95% full,
and cloud tiering is introduced, applying an 80% threshold results in 15 TB of data movement.
Based on network speeds and other factors, the data movement can take a significant
amount of time, almost 2 days over a 1 Gbps link. In cases such as this, it might be best to
gradually reduce the threshold over time to avoid the long initial migrations.

Ad hoc file migration

At any time, you can migrate a specific set of files or directories, such as an archive or a
content repository, to a cloud provider by using a policy file and the mmapplypolicy command.
Example 2 shows how a policy can be used to migrate PDF files in an archive directory.

Example 2 Migrating PDF files in an archive directory

/* Define a list of files to exclude from migration. The .mcstore directory should */
/* not be migrated /*
define(
exclude_list,
(
FALSE
OR PATH_NAME LIKE '/gpfs/.mcstore/%'
)
)

/* Define a cloud pool, using the 'mmcloudgateway' command to perform migrations to */

/* the pool */
RULE EXTERNAL POOL 'cloudpool' EXEC '/usr/lpp/mmfs/bin/mmcloudgateway files' OPTS '-F'
ESCAPE '% -'

/* Migrate files from the system pool to the cloud storage pool */
RULE 'archivemigrate' MIGRATE FROM POOL 'system' TO POOL 'cloudpool' WHERE
(
(
/* Migrate any .pdf files in the application/archive directory */
PATH_NAME LIKE '/gpfs/application/archive/*.pdf%'
)
AND NOT (exclude_list)
)

A policy like this one can be manually run to trigger a bulk migration of these files.

26 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Considerations when running backups
Transparent cloud tiering should not be used as a replacement for a backup strategy. Files
that are migrated with transparent cloud tiering might be lost if accidentally deleted, or during
a complete site failure. File system backups can protect from these types of events.
Transparent cloud tiering can be run along with file system backups for data protection.

When running transparent cloud tiering with file system backups, a file must be backed up
before it is migrated to cloud storage. Failure to back up the file first results in a file recall
when the backup program is run. This can cause unnecessary recall traffic, and add
significant time to the backup run.

To ensure that files are not migrated before being backed up, it is best to define a migration
policy that explicitly excludes files that have been recently modified. The amount of time to
wait to migrate a file should be greater than the frequency of file system backups. For
example, if nightly backups are run, at a minimum, the migration policy should exclude files
that have been modified within the past day.

This setting ensures that they can be backed up before being migrated. In most cases, it is
best to wait longer to avoid recalls if a backup window is missed for some reason. If there is
an issue running backups for a longer time, it might be necessary to disable a migration policy
until the problem with backups has been resolved to avoid many recalls during the next
successful backup.

Excluding recently modified files can be accomplished with a policy statement in the WHERE
clause, such as:
WHERE (DAYS(CURRENT_TIMESTAMP) - DAYS(MODIFICATION_TIME)) > 5

The threshold migration policy that is given above can be extended with this rule, as shown in
Example 3.

Example 3 Extending the use of the threshold migration policy

/* Define a cloud pool, using the 'mmcloudgateway' command to perform migrations to */
/* the pool */
RULE EXTERNAL POOL 'cloudpool' EXEC '/usr/lpp/mmfs/bin/mmcloudgateway files' OPTS '-F'

/* Define a migrate rule, in this example we will migrate from the system pool */
RULE 'MigrateToCloud' MIGRATE FROM POOL 'system'
/* Define the threshold to use, in this example we will migrate when the pool is */
/* 85% full, and attempt to migrate files until the pool is 75% full */
THRESHOLD (85, 75)

/* Next, define a weight for the file to determine which files to migrate first */
/* Here we will use the last access time for the file by subtracting it from the */
/* current time, giving a higher weight to older files */
WEIGHT(CURRENT_TIMESTAMP - ACCESS_TIME)

/* Define that we want to move to the cloud pool */

TO POOL ('cloudpool')
/* We don't want to migrate files that are smaller than 8k in size, because there */
/* are little to no space savings due to file overhead */
WHERE (KB_ALLOCATED > 8)
/* Do not migrate cloud services internal files, which reside in the */
/* .mcstore directory */
AND NOT(PATH_NAME LIKE '%/.mcstore/%'
/* Ignore files that have been modified in the past 5 days */
AND (DAYS(CURRENT_TIMESTAMP) - DAYS(MODIFICATION_TIME)) > 5

27
Bulk file restore
While mmcloudgateway files restore is well suited for restoring individual files or file lists, its
performance is limited because it is only runs as a single execution thread. Moreover, the
restore command requires that the corresponding directory structure in the IBM Spectrum
Scale file system exists before the files in the respective directory are restored. The script
below addresses both limitations, allowing for restoring entire file system trees from the cloud
tier in parallel with a single command.

The ptctrecall.bash script is invoked with the number of parallel processes and the path of
the directory that is to be recovered as arguments. The actual degree of parallelism is limited
by the number of cores available on the machine where ptctrecall.bash is invoked. The
script compares the list of files available in the file system with the list of files stored in the
Cloud Directory to identify the files that need to be restored.

It then splits and formats the corresponding file list so that they can be passed to the
mmcloudgateway files restore command. Directories that contain files are automatically
created by the script enabling the re-creation of lost file system trees. Finally
ptctrecall.bash launches the restore operations in parallel through GNU Parallel.

The ptctrecall.bash script is intended as a starting point for readers who want to be able to
restore data from cloud storage tiers. See Example 4. It can be extended to leverage multiple
cloud service nodes for faster recovery through remote execution, for example with the ssh
command. Moreover, recovery with ptctrecall.bash does not preserve metadata associated
with a directory.

The metadata of a directory could be captured through a find -type d <path> | cpio
pipeline redirected to a file. That file can then be migrated to the cloud tier and retrieved at
recovery time so that directories can be created with the correct metadata information.

Example 4 The ptrecall.bash script

#!/bin/bash

# First arg is the number of processes, second arg is the directory to restore
numprocs=$1
dirpath=$(readlink -f $2)

mkdir /tmp/ptct.$$

# Find files stored in cloud storage that are not found in directory
find $dirpath -type f | sort -u > /tmp/filesindir.$$
/opt/ibm/MCStore/bin/mcstore dblist -p $dirpath -r | sort -u > /tmp/filesincloud.$$
comm -3 /tmp/filesincloud.$$ /tmp/filesindir.$$ > /tmp/filestorestore.$$

# (Re-)create directories that don't exist

/usr/bin/awk 'BEGIN {FS="/";OFS=FS} { $NF=""; print $0 }' /tmp/filestorestore.$$ > /tmp/dirs.$$
sort -u /tmp/dirs.$$ > /tmp/ptct.$$/newdirs
while read -r line
do
if [ ! -d $line ]; then
mkdir $line
fi
done < /tmp/ptct.$$/newdirs
rm /tmp/ptct.$$/newdirs

28 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
# Create chunked file list to distribute restore workload
numlines=$(wc -l < /tmp/filestorestore.$$)
if [ $(( $numlines % $numprocs )) -ne 0 ]; then
chunklines=$(( $numlines / $numprocs + 1 ))
else
chunklines=$(( $numlines / $numprocs ))
fi
split -l $chunklines /tmp/filestorestore.$$ /tmp/ptct.$$/tcchunk

# Create mcstore file lists from chunk files

for i in $(ls /tmp/ptct.$$/tcchunk*)
do
migrationchunk="/tmp/ptct.$$/mcstore.$(basename $i)"
touch migrationchunk
while read -r line
do
echo "filename=$line" >> $migrationchunk
echo "target=$line" >> $migrationchunk
echo "%%" >> $migrationchunk
done < $i
rm -f $i
done

# Launch parallel restore using gnu parallel - max one restore process per core
/usr/local/bin/parallel /opt/ibm/MCStore/bin/mcstore restore -l {} ::: /tmp/ptct.$$/*

# Clean up
rm -f /tmp/filesincloud.$$
rm -f /tmp/filestorestore.$$
rm -rf /tmp/ptct.$$

Interoperability with IBM Spectrum Scale features

For the latest information about interoperability of Cloud Services and other IBM Spectrum
Scale features, see the following resources:
򐂰 IBM Spectrum Scale FAQs
򐂰 Using transparent cloud tiering with other IBM Spectrum Scale features

IBM Spectrum Scale native compression and encryption

IBM Spectrum Scale can natively compress and encrypt files that have been stored to a file
system. Encryption and compression affect data at rest and is transparent to the user.
Transparent cloud tiering migrates and recalls files that have been encrypted or compressed
natively in IBM Spectrum Scale. However, it is worth noting that files are decrypted and
decompressed before being migrated to a cloud storage provider.

In order to ensure at rest encryption of migrated files, the Cloud Service connection can be
configured to encrypt data before sending it to cloud storage. Alternatively, some cloud
storage providers, such as IBM Cloud Object Storage, can be configured to encrypt data at
rest. If a file is recalled from the cloud provider, recalled data is reencrypted using the active
IBM Spectrum Scale encryption policy at the time of the recall.

29
Reconcile
Transparent cloud tiering automatically removes reversioned or deleted files from cloud object
storage. However, the process is not guaranteed to remove files in the event of system
failures or unexpected power loss. The reconcile process will check for such missed files and
absolutely guarantee that they will be deleted, which may be important if you have a
regulatory requirement to guarantee deleted files are destroyed.

The reconcile process takes care of keeping the local file system and its cloud tiered data in
sync by removing such files. The frequency of reconcile (weekly, bi-weekly, or monthly)
should be set depending on how frequently files from the active file system are being
removed, Details of the reconcile process can be found in the “Reconciling files between IBM
Spectrum Scale file system and cloud storage tier” topic at IBM Knowledge Center.

If reconcile is not run periodically, the tiering service might accumulate orphan objects (for
which the corresponding local file from the IBM Spectrum Scale file system is removed) on
the cloud storage. An orphan object is an object in cloud object storage that does not have a
corresponding file in the active IBM Spectrum Scale file system.

Orphaned or previous versions of files can be restored from the cloud tier if they were
accidentally deleted or modified. It can be useful to wait between running reconciles, or to
retain files for several days with the --days-retained flag when running the reconcile process
to preserve these files for some time. See the “Restoring files from the cloud tier” on page 22
for details on restoring previous versions of files.

Monitoring
This section is intended for administrators. It provides guidance about where performance
metrics and events of cloud tiering service can be viewed on the IBM Spectrum Scale GUI,
and how to get an end-to-end capacity view of IBM Spectrum Scale and IBM Cloud Object
Storage by using IBM Spectrum Control™.

Service status
The status of the transparent cloud tiering service and, various entities such as cloud service,
cloud account, cloud storage access point and container pair, on the cloud service node can
be monitored by using the mmcloudgateway service status command to ensure that the
process is running on the cloud service nodes. See the “Transparent cloud tiering status
description” topic in IBM Knowledge Center for details of each status value.

To ensure no disruptions to file access, the cloud service must be running on at least one of
the cloud services nodes at any time. From IBM Spectrum Scale release 5.0.0 onwards,
cloud service is also monitored by IBM Spectrum Scale health monitoring service periodically.
In case of any change in the health state of the service as well as the cloud service entities,
appropriate events are raised and can be monitored through IBM Spectrum Scale GUI.
Details of “Transparent Cloud Tiering events” can be found in IBM Knowledge Center.

The transparent cloud tiering service can also be monitored by using the mmhealth command.
This command provides a single view of multiple IBM Spectrum Scale processes to check the
overall health of a node or set of nodes.

30 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 The CLOUDGATEWAY component in mmhealth (Example 5) shows the status of the transparent
cloud tiering services.

Example 5 Sample mmhealth output

[root@tct-ces1 ~]# mmhealth node show

Node name: tct-ces1

Node status: HEALTHY

Component Status Reasons

-------------------------------------------------------------------
GPFS HEALTHY -
NETWORK HEALTHY -
FILESYSTEM HEALTHY -
DISK HEALTHY -
CES HEALTHY -
CLOUDGATEWAY HEALTHY -
GUI HEALTHY -
PERFMON HEALTHY -

For detailed status of cloud services component, invoke mmhealth node show cloudgateway. It
displays a detailed health view of all of the container pair sets, and any recent events raised.

For additional details, see the “Cloud services: Transparent cloud tiering and Cloud data
sharing” topic in IBM Knowledge Center. It covers monitoring cloud service, administering
files managed by cloud service, recommendations for periodic transparent cloud tiering
service maintenance tasks, and so on.

Cloud connectivity
Connectivity to a cloud storage provider can be verified from any cloud service node by
running the command mmcloudgateway account test. This command verifies connectivity
and credentials on the cloud provider. Verify this information if there are any network or object
storage provider changes, to ensure continued connectivity to the cloud.

Performance Statistics
Transparent cloud tiering performance statistics are available through the IBM Spectrum
Scale GUI, the mmperfmon command, and can be exported to external monitoring tools, such
as Grafana. To view performance statistics of cloud tiering service, it must be enabled in the
IBM Spectrum Scale GUI and monitoring tool. For more information, see these resources:
򐂰 Enabling transparent cloud tiering performance monitoring metrics on the GUI
򐂰 Integrating transparent cloud tiering metrics with performance monitoring tool

The metrics available for transparent cloud tiering include such items as migrate and recall
times, data transferred, and many others. A full “List of performance metrics” is available in
the performance monitoring section of the IBM Spectrum Scale topic in IBM Knowledge
Center.

Each metric has different aggregation levels. Cloud service node, cloud service node group,
and file system can be used to give node level or cluster-wide views of various statistics.

During migrations, you can monitor these metrics in the UI for potential problems.

31
IBM Spectrum Control
IBM Spectrum Control can be used to monitor IBM Spectrum Scale with transparent cloud
tiering and IBM Cloud Object Storage deployments. as shown in Figure 12. IBM Spectrum
Control can be a single point of control in hybrid-cloud deployments using these two products.

Figure 12 IBM Spectrum Control and cloud storage pools

IBM Spectrum Control can list cloud object pools, for example the mcstore pool in Figure 13,
listing the account and cloud containers. In addition, the health of cloud service nodes is
included.

Figure 13 IBM Spectrum Control node status

32 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
 IBM Spectrum Control displays the status of IBM Cloud Object Storage, as shown in
Figure 14. This window provides a single convenient point to view system status.

Figure 14 IBM Spectrum Control and IBM Cloud Object Storage

Security
Cloud services on IBM Spectrum Scale send and receive data from external systems.
Because many of these systems are on a public cloud, data security is extremely important to
consider when configuring a cloud tiering solution. Consider both network security for data
that is in flight, and data security for data that is being stored on the cloud.

Network security
IBM Spectrum Scale network security is described in the IBM Spectrum Scale section of the
IBM Knowledge Center, along with firewall settings that are required to secure the IBM
Spectrum Scale cluster. Cloud service nodes run a service that other nodes in the cluster
communicate with to coordinate data migration and recalls.

By default, this service runs on TCP port 1817. The port can be changed by using the
mmcloudgateway command. Only systems that are a part of the IBM Spectrum Scale cluster
require access to this port, which can be blocked from systems external to the cluster itself.

Cloud services communicate with the external cloud service provider through HTTP or
HTTPS. If you use HTTP, login credentials are sent in plain text. HTTP is susceptible to
malicious users who monitor network traffic to find these credentials. Therefore, use HTTP
only on private cloud systems on secure networks.

To avoid man-in-the-middle attacks, where a malicious user intercept traffic between the
cloud service node and the cloud provider, cloud tiering performs strong checks on any
HTTPS server certificates for authenticity. If you use a private cloud, or a nonstandard
configuration, it might be required to manually update the server certificate by using the
mmcloudgateway service command.

Cloud services initiates TCP connections with the cloud object provider. In many cases, this
process involves connecting through a WAN to a public cloud provider. Under no conditions
do cloud service nodes require inbound connections from the WAN, so all inbound
connections from the WAN should be blocked.

33
Data security
Cloud services offer native encryption to ensure that data is secure when it is at rest on cloud
storage. Data is encrypted by using a secure private key on the cloud service node before
sending it to the external cloud object store. The private key can be managed by using a
secure key server, such as an ISKLM server, or can be configured directly on the cloud
service nodes, where it is stored in a local Java keystore.

Local keys can be used for testing and possibly for some small deployments. A secure server,
such as an ISKLM server, is advised when possible to manage encryption keys. Encryption
keys should be handled carefully, and you should keep a secure backup of the encryption key.
Loss of the key results in the loss of access to any data that has been sent to the cloud
provider.

ISKLM servers also offer key rotation, where a new key can be generated to encrypt new
data. For key rotation, data that has already been migrated by using the old encryption key
will not be reencrypted. For this reason, cloud services store both the old and new encryption
key. The old key is used to decrypt data that has already been migrated to the cloud provider.
The new key is used to encrypt any new data that is being sent to the cloud provider.

Summary
IBM Spectrum Scale transparent cloud tiering allows your existing applications to take
advantage of the flexibility, scalability, and low cost of cloud object storage. Transparent cloud
tiering allows you to migrate data that is accessed less frequently to object storage and recall
data transparently as it is requested by applications.

The IBM Spectrum Scale ILM policy engine provides administrators with a powerful and
flexible tool to manage data being moved to the cloud. End-to-end file validation and available
encryption ensure that data remains secure.

This paper described how transparent cloud tiering can be used to create a hybrid cloud
solution for Active Archives and high-performance workloads such as analytics. It covered
sizing, configuration guidelines, monitoring, and security to help with planning and
implementation. This paper also covered how to configure IBM Cloud Object Storage for use
with transparent cloud tiering, and how IBM Spectrum Control provides integrated
management of a hybrid cloud solution.

34 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Related publications
The publications that are listed in this section are considered suitable for a more detailed
discussion of the topics that are covered in this paper.

IBM Redbooks
The following IBM Redbooks publications provide additional information about the topic in this
document. Some publications referenced in this list might be available in softcopy only.
򐂰 A Deployment Guide for IBM Spectrum Scale Object, REDP-5113
򐂰 IBM Private, Public, and Hybrid Cloud Storage Solutions, REDP-4873
򐂰 IBM Spectrum Scale (formerly GPFS), SG24-8254
򐂰 Introduction Guide to the IBM Elastic Storage Server, REDP-5253

You can search for, view, download, or order these documents and other IBM Redbooks
publications, IBM Redpapers™, Web Docs, draft and additional materials, at the following
website:

ibm.com/redbooks

Other publications
These publications are also relevant as further information sources:
򐂰 IBM Spectrum Scale V5.0 Command and Programming Reference, SC27-9222
򐂰 IBM Spectrum Scale V5.0: Administration Guide, SC27-9220
򐂰 IBM Spectrum Scale V5.0: Concepts, Planning, and Installation Guide, GC27-9219
򐂰 IBM Spectrum Scale V5.0: Problem Determination Guide, SC27-9221
򐂰 IBM Spectrum Scale V5.0: Big Data and Analytics Guide, SC27-9223

Online resources and references

The following resources and references are available online:
򐂰 IBM Client Demonstration Center
https://www.ibm.com/systems/clientcenterdemonstrations

Note: The IBM Client Demonstration Center (for Business Partners, IBMers, and
anyone with an IBMid) provides a catalog of remote demonstrations (video or live
connection) that consists of self-contained material for customer demonstrations of IBM
solutions. Most of the demonstrations are provided with predefined scenarios, and
some also allow for the development of new scenarios. Demonstrations can also be
considered as “ready to use” material for enablement or training.

򐂰 IBM Cloud Object Storage

https://www.ibm.com/cloud/object-storage
򐂰 IBM Elastic Storage Server
https://www.ibm.com/us-en/marketplace/ibm-elastic-storage-server

35
 򐂰 GNU Parallel: The Command-Line Power Tool - O. Tange, February 2011:42-47:
https://www.usenix.org/system/files/login/articles/105438-Tange.pdf
򐂰 IBM Spectrum Scale
https://www.ibm.com/us-en/marketplace/scale-out-file-and-object-storage
򐂰 IBM Spectrum Scale in the IBM Knowledge Center
http://www.ibm.com/support/knowledgecenter/SSFKCN/gpfs_welcome.html
򐂰 IBM Spectrum Scale Overview and Frequently Asked Questions (FAQ)
http://ibm.co/1IKO6PN
򐂰 IBM Spectrum Scale wiki
https://ibm.biz/BdXVxv

Authors
This paper was produced by a team of specialists from around the world working at the IBM
Storage and SDI development lab, Pune India.

Nikhil Khandelwal is a senior engineer with the IBM Spectrum Scale development team. He
has over 15 years of storage experience on NAS, disk, and tape storage systems. He has led
development and worked in various architecture roles. Nikhil is currently a part of the IBM
Spectrum Scale client adoption and cloud teams.

Rob Basham is a Senior Engineer working as a Storage Architect with IBM System Labs. He
has an extensive background working on systems management and storage architecture and
standards such as SCSI and Fibre Channel. Rob is an IBM Master Inventor with many
patents relating to storage and cloud storage. He is currently working as an architect on cloud
storage services including transparent cloud tiering.

Sandeep R. Patil is a Senior Technical Staff Member who works as a Storage Architect with
IBM System Labs. He has over 15 years of extensive product architecture and design
experience. Sandeep is an IBM Master Inventor, an IBM developerWorks® Master Author,
and a member of the IBM Academy of Technology. Sandeep holds a Bachelor of Engineering
(Computer Science) degree from the University of Pune, India.

Anbazhagan Mani is a senior engineer and cloud storage architect. He is an IBM master
inventor and also a member of Technical Experts Council, India. He has over 16 years of
experience working at IBM and has been primarily working on systems management, storage
management, and cloud technologies including OpenStack and IBM Cloud Object Storage.
He is one of the architects of transparent cloud tiering.

Amey Gokhale has over 15 years (9+ years with IBM) of industry experience in various
domains including systems, networking, medical, telecom, and storage. In IBM, he has mainly
led teams working on systems and storage management. Starting with IBM Storage
Configuration Manager, which managed internal LSI RAID controllers in System x, he led a
team that provided storage virtualization services within IBM System Director VMControl,
which met storage requirements in virtualized environments. In his current role, he is a
co-architect of transparent cloud tiering in Spectrum Scale and is leading the ISDL
development team responsible for productization, install, serviceability, and configuration.

36 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Jinesh Shah is a Senior Solution Architect, member of IBM GIST (India) Team. He has
worked on various platforms, datacenter activities, and consulting projects. In his current role,
he supports IBM partners for global RFPs and outsourcing deals as a Storage Pre-sales
consultant. Jinesh has over 19 years of IT experience. He is a science graduate from
University of Mumbai, India.

Kedar Karmarkar is a senior engineer and solution architect with IBM Spectrum Scale
development team. Kedar is part of the IBM Spectrum Scale Client adoption team and was
the IBM Storwize® V7000 Unified Level 3 support lead in his prior role at IBM. He has over 20
years of infrastructure software, storage development experience in management, and
architect roles. He has led development of NAS Storage, Block level virtualization, replication,
systems, and storage management products. Kedar has a Bachelor of Engineering
(Computer Science) degree from the University of Pune, India.

Donald Mathisen is an IT professional with over 25 years of experience who is currently a

Senior Managing Consultant for Systems Lab Services. He is IBM Certified for Systems
Storage, an IBM Level 2 Certified IT Specialist, and a The Open Group Master Certified IT
Specialist. He has held several roles of increasing responsibility with previous companies
before coming to IBM. Don started with IBM in Strategic Outsourcing as an Advisory
Technical Services Professional in March of 2006. He is the lead person in Systems Lab
Services for IBM Object Storage, transparent cloud tiering, Elastic Storage Server, IBM XIV®,
and IBM Spectrum Accelerate™ platforms. In addition, he has led many cross-functional
worldwide IBM teams on customer facing engagements. He has provided worldwide
education and training for services delivery on such products as XIV, FlashSystems, IBM
Spectrum Accelerate in the Cloud, and IBM Spectrum Accelerate. Don holds a BA in
Computer Science from Mt. Saint Mary College in Newburgh, NY.

Larry Coyne is a Project Leader at the International Technical Support Organization, Tucson
Arizona center. He has 35 years of IBM experience with 23 in IBM storage software
management. He holds degrees in Software Engineering from the University of Texas at
El Paso and Project Management from George Washington University. His areas of expertise
include client relationship management, quality assurance, development management, and
support management for IBM storage software.

Arend Dittmer is a Solution Architect for Software Defined Infrastructure. He has over fifteen
years of experience with scalable Linux based solutions in enterprise environments. His
areas of interest include virtualization, resource management, big data, and high
performance parallel file systems. Arend holds a patent for the dynamic configuration of
hypervisors for portable virtual machines.

Thanks to the following people for their contributions to this project:

Tito Hernandez
IBM Cloud

Keith D. Ball
Michael Daubman
Scott Fadden
Pallavi Galgali
Arati Ranade
Harald Seipp
Frederick Stock
Christine Telford
Carl Zetie
IBM Systems

37
 Lisa Shepard
IBM Global Technology Services®

Now you can become a published author, too

Here’s an opportunity to spotlight your skills, grow your career, and become a published
author—all at the same time. Join an ITSO residency project and help write a book in your
area of expertise, while honing your experience using leading-edge technologies. Your efforts
will help to increase product acceptance and customer satisfaction, as you expand your
network of technical contacts and relationships. Residencies run two - six weeks in length,
and you can participate either in person or as a remote resident working from your home
base.

Find out more about the residency program, browse the residency index, and apply online:
ibm.com/redbooks/residencies.html

Stay connected to IBM Redbooks

򐂰 Find us on Facebook:
http://www.facebook.com/IBMRedbooks
򐂰 Follow us on Twitter:
http://twitter.com/ibmredbooks
򐂰 Look for us on LinkedIn:
http://www.linkedin.com/groups?home=&gid=2130806
򐂰 Explore new Redbooks publications, residencies, and workshops with the IBM Redbooks
weekly newsletter:
https://www.redbooks.ibm.com/Redbooks.nsf/subscribe?OpenForm
򐂰 Stay current on recent Redbooks publications with RSS Feeds:
http://www.redbooks.ibm.com/rss.html

38 Enabling Hybrid Cloud Storage for IBM Spectrum Scale Using Transparent Cloud Tiering
Back cover

REDP-5411-01

ISBN 0738456861

Printed in U.S.A.

®
ibm.com/redbooks