These materials are © 2015 John Wiley & Sons, Inc.

Any dissemination, distribution, or unauthorized use is strictly prohibited.

 Flash Array
Deployment

IBM Limited Edition

By Neal Ekker

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Flash Array Deployment For Dummies®, IBM Limited Edition
Published by
John Wiley & Sons, Inc.
111 River St.
Hoboken, NJ 07030‐5774
www.wiley.com
Copyright © 2015 by John Wiley & Sons, Inc.
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise,
except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without the
prior written permission of the Publisher. Requests to the Publisher for permission should be
addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ
07030, (201) 748‐6011, fax (201) 748‐6008, or online at http://www.wiley.com/go/permissions.
Trademarks: Wiley, For Dummies, the Dummies Man logo, The Dummies Way, Dummies.com,
Making Everything Easier, and related trade dress are trademarks or registered trademarks of John
Wiley & Sons, Inc. and/or its affiliates in the United States and other countries, and may not be used
without written permission. IBM, the IBM logo, FlashSystem, FlashCore technology, Real-time
Compression, and Variable Stripe RAID are trademarks or registered trademarks of International
Business Machines Corporation. All other trademarks are the property of their respective owners.
John Wiley & Sons, Inc., is not associated with any product or vendor mentioned in this book.

LIMIT OF LIABILITY/DISCLAIMER OF WARRANTY: THE PUBLISHER AND THE AUTHOR MAKE

NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE ACCURACY OR
COMPLETENESS OF THE CONTENTS OF THIS WORK AND SPECIFICALLY DISCLAIM ALL
WARRANTIES, INCLUDING WITHOUT LIMITATION WARRANTIES OF FITNESS FOR A
PARTICULAR PURPOSE. NO WARRANTY MAY BE CREATED OR EXTENDED BY SALES OR
PROMOTIONAL MATERIALS. THE ADVICE AND STRATEGIES CONTAINED HEREIN MAY NOT BE
SUITABLE FOR EVERY SITUATION. THIS WORK IS SOLD WITH THE UNDERSTANDING THAT
THE PUBLISHER IS NOT ENGAGED IN RENDERING LEGAL, ACCOUNTING, OR OTHER
PROFESSIONAL SERVICES. IF PROFESSIONAL ASSISTANCE IS REQUIRED, THE SERVICES OF A
COMPETENT PROFESSIONAL PERSON SHOULD BE SOUGHT. NEITHER THE PUBLISHER NOR
THE AUTHOR SHALL BE LIABLE FOR DAMAGES ARISING HEREFROM. THE FACT THAT AN
ORGANIZATION OR WEBSITE IS REFERRED TO IN THIS WORK AS A CITATION AND/OR A
POTENTIAL SOURCE OF FURTHER INFORMATION DOES NOT MEAN THAT THE AUTHOR OR
THE PUBLISHER ENDORSES THE INFORMATION THE ORGANIZATION OR WEBSITE MAY
PROVIDE OR RECOMMENDATIONS IT MAY MAKE. FURTHER, READERS SHOULD BE AWARE
THAT INTERNET WEBSITES LISTED IN THIS WORK MAY HAVE CHANGED OR DISAPPEARED
BETWEEN WHEN THIS WORK WAS WRITTEN AND WHEN IT IS READ.

For general information on our other products and services, or how to create a custom For Dummies
book for your business or organization, please contact our Business Development Department in the
U.S. at 877‐409‐4177, contact [email protected], or visit www.wiley.com/go/custompub.
For information about licensing the For Dummies brand for products or services, contact
BrandedRights&[email protected].
ISBN: 978‐1‐119‐10219‐9 (pbk); ISBN: 978‐1‐119‐10208‐3 (ebk)
Manufactured in the United States of America
10 9 8 7 6 5 4 3 2 1

Publisher’s Acknowledgments
Some of the people who helped bring this book to market include the following:

Project Editor: Carrie A. Johnson Business Development Representative:

Editorial Manager: Rev Mengle Sue Blessing
Acquisitions Editor: Steve Hayes Key IBM contributors: Matthew S. Key,
Philip A. Clark, Arati V. Vora

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Table of Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
About This Book......................................................................... 1
Icons Used in This Book............................................................. 2

Chapter 1: Learning Why Data Storage

Performance Matters. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Speed Matters Online................................................................. 3
eCommerce........................................................................ 5
Big Data and analytics...................................................... 6
Financial services............................................................. 7
Cloud, mobile, and social engagement.......................... 7
Performance Drives Value......................................................... 8

Chapter 2: Getting to Know Flash Storage Systems . . . 11

Solid State Drives...................................................................... 12
SANs and Storage Arrays......................................................... 14
PCIe Cards.................................................................................. 16
Solid State Arrays...................................................................... 17

Chapter 3: Choosing Flash Storage Arrays. . . . . . . . . . . 21

Storage Analysis........................................................................ 21
Flash Controllers....................................................................... 23
Looking at the Challenges of SSDs.......................................... 25
Understanding the Liabilities of PCIe Cards.......................... 26
Establishing the Advantages of Flash Arrays........................ 27
IBM FlashSystem....................................................................... 28

Chapter 4: Exploring Deployment Designs

with IBM FlashSystem. . . . . . . . . . . . . . . . . . . . . . . . . . 31
Deploying Direct Attached Storage with IBM FlashSystem.... 32
Connecting Systems with a SAN-Attached
Architecture........................................................................... 35

Chapter 5: Implementing the Future with

Virtualized Storage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Storage Pooling and Tiering.................................................... 38
Data Protection......................................................................... 40
Capacity Optimization.............................................................. 43

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
iv Flash Array Deployment For Dummies, IBM Limited Edition_______

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Introduction
O rganizations ranging from city governments, through
movie production companies and medical research
institutions, to car manufacturers are realizing that how well
their information technology (IT) performs strongly affects
how well their business performs. Insightful executives and
IT managers understand that their data storage systems play
a crucial role in how well their information technology helps
them achieve important objectives such as faster decision
making, better customer service, or a smaller data center
budget. Solid state storage made from NAND flash memory
chips has evolved in terms of cost, performance, and reli­
ability to the point where many organizations are seriously
considering its use to replace inefficient, unacceptably slow
mechanical spinning disk systems. This accelerating trend
has led enterprises to ask some natural questions: When
should flash be used? Which flash solution is best for each
particular use case? And how can I make it a successful, cost‐
effective part of my data center? These are the questions I
answer (especially the last one) in Flash Array Deployment For
Dummies, IBM Limited Edition.

About This Book

If you’re a decision maker in an enterprise determined to make
more, spend less, and move faster, this book is for you. Flash
Array Deployment For Dummies, IBM Limited Edition, tackles the
data storage challenges of “enterprises” — commercial, scien­
tific, and governmental organizations. It does not address “con­
sumer” data storage issues, such as those faced by privately
owned PC, smartphone, laptop, and iPad devices, and so on.

Chapters 1 and 2 of this book are most helpful to decision

makers. In these chapters, I introduce some of the data storage‐
related problems you see that have led you to consider flash
storage, discuss why you may choose to solve these prob­
lems with flash storage, and highlight some benefits if you do.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
2 Flash Array Deployment For Dummies, IBM Limited Edition _______

Also in the early chapters, I introduce the various types of

flash storage and explain what they’re used for and who cur­
rently uses them. As the name implies, this book is ultimately
about flash storage arrays — flash devices that can stand alone
and are used most often in data center environments where
multiple computers (servers) can access or share the same
storage solution. Flash arrays offer good solutions to the major­
ity of storage challenges you may be experiencing in your data
center.

Chapters 3 and 4 provide the most current thinking about

what you should do as the responsible manager or technician
if you are assigned the task of actually implementing a flash
storage solution. Of course, this information can be invalu­
able to those working on the data center floor. But it may also
prove helpful to IT decision makers because how effectively
your flash storage solution is deployed, configured, and
operated will play a large role in the return you see in your
flash storage investment — a matter dear to the hearts, and
careers, of IT decision makers.

Icons Used in This Book

You’ll find several icons in the margins of this book. Here’s
what they mean.

A Tip is a suggestion or a recommendation. It usually points

out a quick and easy way to get things done or provides a
handy piece of extra information.

The Warning icon alerts you to conditions that require

extra care and thinking. For example, you don’t want to omit
critical steps in evaluating your needs and planning your
implementation.

Anything that has a Remember icon is something that you

want to keep in mind.

Technical Stuff contains information that’s interesting and

useful but not vital to understanding flash array deployment.
Info here may include a brief history of a principle, the earliest
practitioners, or the origin of a word. It also showcases tech­
nical points. You can either read these or skip over them.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Chapter 1

Learning Why Data Storage

Performance Matters
In This Chapter
▶▶Learning why storage speed matters to online applications
▶▶Seeing how storage speed accelerates traditional IT environments

I nformation technology (IT) isn’t an end in itself; it’s a

means to solving certain business problems or enhancing
business opportunities. A data storage solution that makes
life better in the data center but doesn’t contribute positively
to your organization’s success isn’t really a solution at all. So,
the first order of business in learning more about flash array
deployments is to connect your data storage to your business
challenges.

Speed Matters Online

Let’s say that you’re an enterprise that accomplishes at least
some of your business activities online. Many organizations
fall into this group — everyone from retailers selling products
directly to customers through the Internet, to banks and other
financial institutions offering services online, to scientific
organizations sharing research information with c ­ olleagues
around the world. Yet the Internet, at its most basic level,
is just a collection of computers exchanging digital bits —
ones and zeroes, pulses of high and low electrical voltage.
Computers connected to each other and exchanging informa-
tion are known as networks. Such collections or networks of
computers have grown to be very complex and powerful over
the years in terms of how much digital information they can

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
4 Flash Array Deployment For Dummies, IBM Limited Edition _______

transfer over various connective media, such as metal wires,

optical fibers, and even various frequencies of electromag-
netic waves (wireless), and how well they can manage the
streams of information zooming between them. They need to
be powerful because the amount of data that you may want to
transfer across various digital networks is rapidly increasing.

Around 2.5 quintillion bytes (a unit of digital information most

commonly consisting of eight bits) of new data is created
every day — and by 2017, IBM predicts that data volumes will
grow by another 800 percent.

In order for computer networks to manage and transfer this

ever‐growing enormous amount of data successfully, they must
become ever faster. Data is constantly stored on and retrieved
from many of the computers or “nodes” involved in networks,
and the speed of data storage directly affects the overall per-
formance of things you want to do using computer networks.

Simple transfers of information from one computer to another

are just the tip of the iceberg of reasons why people use net-
works. They often access computer software over the Internet
and other types of networks, such as Local Area Networks
(LAN) within or controlled by organizations or geographically
wider LANs called Wide Area Networks (WAN). Network‐
connected applications allow you to share photos with your
loved ones, transfer money from one account to another,
and even hold business meetings with colleagues around the
world. But none of this happens effectively without fast data
storage and retrieval.

Essentially all business, government, and research activities

in the modern world use computer applications as founda-
tional tools. As the volume of data increases, the applications
on which you depend must grow ever faster. In order for
applications to perform more work for you in shorter time
frames, they must store or write data to storage devices and
retrieve or read this data in the least amount of time possible.
The amount of time taken for data storage round trips, essen-
tially the storage response times, is known as storage latency.

One of the most important limits or throttles on an applica-

tion’s ability to perform useful work is the storage latency
of the computer system on which the application runs or is
hosted. Storage latency is a central and crucial concept within
this book.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 1: Learning Why Data Storage Performance Matters 5
There is another kind of latency within digital systems —
network latency. When applications operate partly or mostly
over networks, network latency is added to the computational
and storage latencies that exist where the application is actu-
ally running, so network‐based applications must address
both their own local latencies and the network latency as well.
Dealing with the local storage latency is bad enough; when
you add network latency, the challenges multiply.

The applications employed as vital tools by your enterprise

are limited in their performance by the latency and other
­performance‐related characteristics of the associated ­computer
system’s data storage devices and designs or architectures.
Addressing this one issue itself has spawned large and thriving
industries within the world of information technology. Then, if
your organization’s activities utilize computer n
­ etworks — the
Internet or your own LANs or WAN — the issues associated
with application performance, storage latency, and the need
for speed grow even more thorny and challenging.

But where there is challenge, there may also be opportunity.

Whole new industries and sectors of economic endeavor have
been created by the advent of networked computing, and
most others have been transformed or at least significantly
affected by these technologies. In almost every case, fast data
storage is a fundamental requirement for success.

eCommerce
Online retail activities, or what is called eCommerce, has
become a driving force of global economics. And eCommerce
provides an excellent example of why data storage perfor-
mance matters. Business to customer sales facilitated by the
Internet surpassed $1 trillion several years ago, according to
IBM research, and will soon account for over 5 percent of all
worldwide economic activity. In addition to creating entirely
new business models, eCommerce also competes directly with
traditional brick and mortar stores. For example, currently in
the United States 70 percent of consumers experience their
first interaction with a brand online, and within a few years
50 percent of all retail dollars spent in the United States will be
partly or entirely transacted digitally.

Online shoppers don’t want to wait for information to make

their buying decisions; they demand rapid responses. Rich

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
6 Flash Array Deployment For Dummies, IBM Limited Edition _______

and complex web pages with dynamic content take longer to

load, especially on mobile devices, and that creates a funda-
mental challenge for eCommerce providers.

The influence of eCommerce is growing in the overall mar-

ketplace by about 20 percent every year. As utilization and
market opportunity escalate, the IT backbone supporting
eCommerce becomes a critical path component to success.
Retail websites that deliver information about products and
interact with customers quickly and reliably usually increase
both market share and profitability.

Big Data and analytics

Enterprises today collect tremendous volumes of data that’s
generated by a wide range of sources often at extreme veloci-
ties. These massive data sets are called Big Data. Discovering
and communicating meaningful patterns in these large collec-
tions of data is called Big Data Analytics.

For all businesses, data itself is one of your most valuable

assets, and Big Data Analytics may already be one of the most
powerful new tools you use to gain competitive advantage,
increase sales, and protect your business from fraud. But,
the near real‐time analysis and response velocities of Big
Data Analytics require a storage environment with the lowest
possible system latency. And the rapid transfer of enormous
data volumes requires extraordinary storage bandwidth.
Therefore, storage performance truly matters in any enter-
prise hoping to harness the benefits of Big Data.

Science needs performance too

Non‐profit enterprises also mine annually, or around 500 quintillion
Big Data for value. Take the Large (5×1020) bytes per day — almost
Hadron Collider (LHC) for example, 200 times higher than all other data
outside Cern, Switzerland. LHC sources in the world combined.
experiments involve about 150 million From analyzing that data, LHC scien-
sensors delivering data 40 million tists found glimmers of evidence for
times per second. The data flow the existence of the Higgs Boson or
could exceed 150 million petabytes “God particle.”

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 1: Learning Why Data Storage Performance Matters 7
Financial services
The financial services industry offers another good example of
why storage performance matters. Few industries have been
so affected and accelerated by the Internet as the financial ser-
vices sector, especially equities trading. Core banking systems
are getting faster as they turn from pure systems of record to
systems of customer engagement, with new online and mobile
access rates increasing dramatically. On the securities side
of the house, equities sales occur in milliseconds now days.
Throughout the financial sector these trends lead to fierce
competition where the performance of IT infrastructure makes
the difference between the firms that capture market share
and profits and those that don’t. System latency and scalability
are of critical importance to applications in this environment.
Beyond operational transaction processing, risk and market
assessment requirements of financial services enterprises
have also fostered the industry‐wide adoption of online analyt-
ical processing (OLAP) tools, further fueling the requirement
for very fast IT systems and high‐performance data storage.

Cloud, mobile, and social

engagement
The future of online enterprise is a wild new world. The Internet
has not only transformed traditional businesses, but also it’s
fostered the creation of entirely new industries, inspired new
avenues for theft and crime, and even enabled a new model
for delivering compute services themselves. This exploding
new world of commerce and interaction, legal and otherwise,
drives an arms race of new data storage technologies and
solutions, all based on the ever‐accelerating need for speed.

Mobile computing and online social engagement are two

of these entirely new enterprises spawned by the Internet.
Literally, they’re already profoundly changing the arena of
global business and society. Proliferating mobile technology
and the spread of social business are empowering people with
knowledge, enriching them through networks, and changing
their expectations. For example, 57 percent of companies now
expect to devote more than a quarter of their IT spending
to mobile and social systems of engagement by 2016, nearly
twice the levels of 2013.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
8 Flash Array Deployment For Dummies, IBM Limited Edition _______

At the same time that industries and professions are being

remade by the Internet, the IT infrastructure of the world is
being transformed by the emergence of Cloud computing.
In all Cloud delivery models the IT infrastructure challenges
related to data storage are similar. Most importantly, because
applications and functionality delivered through a Cloud
model come to end‐users through networks, local or Internet
or both, system latency is a critical issue. Overall response
time includes both network latency and response delays
­generated at the compute source. Networks are growing
ever faster, which shifts much of the focus on reducing
latency to the data center itself and from there directly to
the storage systems. This is why flash array deployments in
Cloud and other network‐centric environments are escalat-
ing, because only high‐performance data storage can enable
the future.

IBM estimates that by 2016, more than one‐fourth of the

world’s applications will be available in the Cloud, and
85 ­percent of new software is now being built for Cloud
­compute environments. The delivery of IT as online services
is creating new business models that are generating a market
expected to reach $250 billion in 2015.

Performance Drives Value

Not all business, governmental, and scientific activity happens
online. If 5 percent of worldwide economic activity is facili-
tated by the Internet, then 95 percent is not. This suggests
there’s a lot of data processing that isn’t network‐enabled
and instead happening locally within the physical walls of
enterprises.

Does this “locally” occurring computer activity need fast data

storage? And do the organizations that depend on computer
programs as crucial tools in their operations really benefit
much from revved up IT infrastructure? In fact, it’s easy to
show that both are very true.

Databases offer the most widely applicable example in the

“local processing” category. Databases and database manage-
ment systems (DBMS) have been around since the dawn of
the information age.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 1: Learning Why Data Storage Performance Matters 9
A database is an organized collection of data typically used
to model aspects of reality in a way that supports processes
that require information, for example, modelling the availabil-
ity of rooms in hotels in a way that supports finding a hotel
with vacancies. DBMSs are computer software applications
that interact with users, other applications, and the database
itself to capture and analyze data. A general‐purpose DBMS is
designed to allow the definition, creation, querying, update,
and administration of databases.

For example, if at work you use applications for financial

accounting, to manage employee records, to provide cus-
tomer service, or to track parts inventories or product sup-
plies, you use applications that rely heavily on databases.
Because databases are involved with the majority of appli-
cations, and data processing is involved in the majority of
economic activity on the planet, it’s not a stretch to suggest
that the performance of databases affects and influences the
activities of business, government, and science more than
almost any other information technology.

Do databases benefit from storage that’s faster than tradi-

tional spinning disks? A recent study conducted by analysts
at Wikibon suggests that overall IT costs can be dramatically
lowered by replacing conventional disk‐based storage with
much higher performance storage:

✓✓54 percent lower overall IT infrastructure cost

✓✓94 percent less administration and operational support
outlays
✓✓76 percent reduction in environmental (power/space)
expenses
✓✓52 percent lower software costs

But how did we leap from exploring how databases can

perform better with faster storage to lowering the costs
associated with the entire data center? It turns out that
the performance of your data storage system dramatically
impacts the costs of deriving value from your information in a
number of ways:

✓✓When you do more work for the same cost, the expense
per unit of work goes down. Faster storage, with lower
latency, enables databases to respond more swiftly to
each request for data from users or applications.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
10 Flash Array Deployment For Dummies, IBM Limited Edition _______

✓✓Over the decades, most IT components have relentlessly

grown faster and more powerful, but traditional spinning
hard disk drives have not, so the speed of data proces-
sors, often called CPUs (central processing unit), is now
much greater than disk‐based storage — on the order of
10,000 times faster (GHz/ms). This means that proces-
sors hosting applications may spin idly through many
cycles waiting for a data request as it travels from CPU
to database to storage and back again. In fact, utilization
rates can hover below 10 percent in many traditional
data centers. But add faster storage, and CPU utilization
rates can shoot up much higher, depending on how well
other system components are optimized. Higher CPU
utilization rates create efficiencies all through IT sys-
tems. Fewer servers can be used to accomplish the same
amount of work. And less software may be needed, run-
ning on those servers.
✓✓It turns out that faster storage isn’t mechanical storage,
and with no moving parts, only electrons, it consumes
much less electricity. This also means it throws off pro-
portionately less heat, which translates into less air cool-
ing needed in the data center. Over the past few years,
environmental costs such as power and HVAC, and even
the value of data center floor space, have risen dramati-
cally. They now figure prominently into any accurate
assessment of enterprise data storage costs.
✓✓Mechanical components, such as spinning disk drives,
tend to break down or wear out faster than electronic
circuitry. This leads to many more repair and reconfig-
ure episodes for database and system administrators.
Plus, the systems designed to mitigate the slow perfor-
mance of traditional storage often grow very complex
and demand plenty of attention. It’s easy to see that a
very wide range of enterprises — from retail websites,
through banks and stock traders, to staff managing per-
sonnel records and product inventories in businesses of
all sizes — can benefit from and in fact are demanding
faster data storage performance. And it’s also clear that
faster storage isn’t based on spinning mechanical disks.

Even with the advent of the Internet, mobile apps, social

engagement, and Cloud computing, databases are still
involved in the vast majority of data processing. It’s easy to
see that the benefits they derive from high‐performance stor-
age will drive flash array deployments for years into the future.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Chapter 2

Getting to Know Flash

Storage Systems
In This Chapter
▶▶Understanding solid state drives
▶▶Discovering storage arrays and storage area networks
▶▶Using PCIe cards
▶▶Defining solid state arrays

C hapter 1 demonstrated that faster data storage for

enterprises of all types and sizes is a matter of cost,
productivity, and competitive advantage. After you make the
decision to implement faster data storage for your enterprise,
you’re ready to move to the second step in the process of
flash array deployment, which is learning about what higher
performance data storage technologies are currently available
to you. And that’s what I cover in this chapter.

Essentially, there’s only one viable option right now — data

storage made from integrated circuits instead of spinning
disks. For decades it has been known as solid state storage.
Originally, solid state storage consisted of random access
memory (RAM) chips aggregated into large integrated groups
or arrays. These devices were used as massive temporary
holding places for data, called buffers or caches. Because RAM
loses data when the power goes off, these devices relied on
other components such as redundant power supplies and
even batteries within their deployment environments to
prevent data loss. Various solid state storage devices made
from DRAM (dynamic RAM, the current version of RAM) are
still available, though they’re very expensive. They’re used in

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
12 Flash Array Deployment For Dummies, IBM Limited Edition _______

unique situations where the lowest possible storage latency is

demanded, no matter the cost.

In the past ten years, solid state storage made from a type
of integrated circuitry called flash memory has become very
popular in consumer electronics because a chip the size of
your thumbnail can hold quite a lot of data, even when you
turn off the device or the battery goes dead. The use of flash
memory chips in a wide and ever‐growing spectrum of con-
sumer products has driven their cost steadily lower over the
past decade while spurring plenty of innovative engineering.
As the cost has fallen and the capabilities and endurance
have dramatically risen, flash has become viable for the more
demanding IT environments found in modern enterprises.

Introduced by Toshiba in 1984, flash memory cells are made

of “floating gate” transistors. NAND flash memory chips are
composed of literally millions of flash cells and form the basis
of devices built for storing the data generated by business,
government, academic, medical, and scientific enterprises
of all types from around the globe. When you compare the
operational expenses (electricity, cooling, floor space), man-
agement costs, server and software outlays, and performance
value of conventional disk systems and flash storage, you
see that even though until recently the purchase price or
dollars per gigabyte ($/GB) of flash devices for enterprise
use has been considerably higher, all the other costs can be
quite a bit lower, making the overall costs much more equiva-
lent. And now that the $/GB are converging as well, the total
cost of ownership (TCO) of the two storage types is tipping
toward flash. Because of this, IT industry analysts predict that
deployments of flash data storage solutions for enterprise use
cases will dramatically rise in the next few years.

Solid State Drives

In the early 1990s, following the invention of flash memory,
a new kind of solid state storage product evolved. Because
disk‐based systems were the most widely used enterprise
storage solutions, only storage products that could operate in
the hard disk drive bays of servers or the disk enclosures of
enterprise storage arrays were practical. And the solid state
drive (SSD) was born.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������ Chapter 2: Getting to Know Flash Storage Systems 13
For nearly 20 years, the term solid state disk was used by most
industry insiders to refer to any solid state storage device, no
matter its shape (form factor), what it was made from, or how it
was used. In the past few years, as the industry has grown and
products have proliferated, the term solid state disk has fallen
out of use and now SSD refers specifically to solid state storage
products with hard disk drive form factors that interface with
storage systems by using industry standard hard disk drive
software or protocols. Such a device can be seen in Figure 2-1.

Figure 2-1: An example of an SSD.

SSDs are products of convenience, cost, and trade‐offs.

Because they’re designed to connect to storage systems just
like traditional disk drives do, they offer a convenient way
for enterprises to add some solid state storage performance
to conventional disk‐dominated environments. As the gap
between the speed and performance of CPUs and disk storage
grew steadily wider over the years, enterprises had stronger
motivation to look for something that could practically inte-
grate with their existing storage systems but provide more
inputs/outputs per second (IOPS) and lower latency than hard
disk drives. SSDs filled the bill.

Additionally, SSDs offer another powerful advantage — the

cost of a unit is considerably lower than other solid state
storage devices. It’s important to note that this doesn’t neces-
sarily mean that the cost per usable storage capacity or the
cost per application transaction capability is lower for SSDs.
A useful analogy might be that of a Chevy pickup versus a
semi‐truck and trailer. It’s quite possible that because the
semi can haul a load of many tons whereas the pickup can
barely haul one ton, the cost per unit of hauling capacity
could actually be lower with the semi. But buying a semi costs
much more than buying a Chevy pickup.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
14 Flash Array Deployment For Dummies, IBM Limited Edition _______

So, in all the applications where a Chevy will do just fine, SSDs
have flourished. On the consumer side, SSDs will fit in your
PC or your laptop without your having to buy added equip-
ment or software to handle them, for the most part. The same
is true with both enterprise servers and conventional storage
systems. In fact, in the past ten years most enterprise operat-
ing systems, virtualization software, and storage array control-
lers have been upgraded to handle SSDs, with varying degrees
of effectiveness. Thanks to their convenience and cost per
unit, SSDs have maintained their rank over the years as the
hottest selling solid state storage devices.

SANs and Storage Arrays

An advantage of SSDs is their ease of deployment in conven-
tional enterprise storage environments — both in servers
and in the large collections of hard disk drives known as
enterprise storage arrays that are deployed in Storage Area
Networks (SAN). A common SAN is shown in Figure 2-2.

A SAN is the standard way in today’s data centers to share a

storage resource such as a storage array with multiple serv-
ers. A SAN is created by networking the servers by way of
Fibre Channel or other connectivity through switches to one
or more storage devices. Each storage device could be a large
enterprise storage array with several, dozens, or even scores
of individual SSDs involved.

Figure 2-2: A common SAN.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������ Chapter 2: Getting to Know Flash Storage Systems 15
An enterprise storage array is a group of integrated hard disk
drives or other storage media devices uses a computer known
as a controller to manage collective activities. Figure 2-3 shows
the disk enclosures, controllers, network switches, and other
related hardware of an enterprise storage array all housed in
a single cabinet.

Figure 2-3: An enterprise storage array with all its components in a single
cabinet.

Your SAN could be composed of multiple storage arrays and

other storage devices, such as tape drives and even optical/
CD drives often used for data archiving purposes, as well as
flash arrays. All these storage devices are connected using an
appropriate networking technology such as Fibre Channel or
Ethernet and then made available to your various application
hosts/servers on the other side of a network switch.

SANs have proven to be powerful and popular storage

designs, or architectures, for decades, but they do introduce
network latency. Every request for data made by an applica-
tion must travel away from the CPU, out of the server enclo-
sure, through Fibre Channel or other networking, through a
switch(s), into a storage array controller, round and about

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
16 Flash Array Deployment For Dummies, IBM Limited Edition _______

through the storage array software and hardware, into the

individual SSD where the data actually resides — then make
the entire return journey.

PCIe Cards
By the early 2000s, enough applications were becoming
­storage latency sensitive that finding some alternatives began
to look like an excellent business venture. Engineers explored
ways to avoid the network latency incurred by traditional
SANs, and soon the Peripheral Component Interconnect
Express (PCIe) card was born (see Figure 2-4). In less than ten
years, this technology has become one of the most successful
solid state storage devices in the marketplace.

Most servers now include PCIe high‐speed connections as

part of their internal architectures. Integrated circuit boards
or cards of various physical sizes can connect directly into
the main server circuitry or bus through slots with certain
numbers of connection pins. At first, PCIe cards were only
that — boards or cards with PCIe connections that held
large onboard flash chip arrays. PCIe cards install directly
into the server enclosures, so they eliminate SAN network
latency. Systems administrators load software that works
with the operating system to manage the PCIe cards and
create substantial pools of flash‐based storage that can be
used to accelerate the performance of latency‐sensitive
applications.

Figure 2-4: A typical PCIe card.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������ Chapter 2: Getting to Know Flash Storage Systems 17
Multiple PCIe cards can be installed in a server, depending on
how many slots are available. Over time, a wide range of soft-
ware and alternative hardware configurations have been devel-
oped to address many different IT infrastructure requirements
and challenges. They all provide lower latency than traditional
SSDs, especially SSDs deployed in SANs, and yet they also offer
a similar advantage enjoyed by SSDs — comparatively lower
unit prices.

Though PCIe configurations offer almost all the options of

other solid state storage devices, most often they’re deployed
to create large pools of in‐server storage that isn’t quite as
fast as DRAM but much faster than disk‐based storage, no
matter where the disks are physically located. The pool or
cache of very fast storage is managed by software that moni-
tors the activity of the data sets used by the application(s)
hosted on the server. The most active data, which will be the
data that can most benefit from ultra‐low latency, is copied
to the PCIe card, and the application then reads it from there,
not from other, slower storage.

The PCIe model of data storage is sometimes described as

server-centric application acceleration. It can offer a range of
benefits. These flash products cost much less than today’s
DRAM while offering memory‐like performance. Because the
space inside a server enclosure limits their size, you can save
on purchase or capital expenses over some other solid state
storage products, and because they’re made from flash chips,
you save operational costs. They can be targeted at accelerat-
ing a single mission‐critical application. And certainly PCIe
cards solve the latency challenge better than any other solid
state storage devices.

Their sales in the marketplace have skyrocketed over the

past seven years. Nonetheless, both PCIe‐based storage and
SSDs still compete with the original type of solid state storage
device — the standalone appliance.

Solid State Arrays

Before SSDs were invented and long before any PCIe card
existed, there were standalone solid state storage appliances.
Contrary to some predictions of a few years ago, deployments
of flash‐based solid state arrays have been substantially

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
18 Flash Array Deployment For Dummies, IBM Limited Edition _______

increasing, rather than declining due to the competition

from PCIe cards and SSDs. Figure 2-5 shows you one of the
current solid state storage array products available in the
marketplace.

Figure 2-5: Example of a current solid state storage array.

Although solid state arrays (SSA) were once entirely com-

posed of RAM, now those are rare and built for special-use
cases. Instead, the vast majority of SSAs these days are flash‐
based, and they come in a wide variety of shapes and sizes.
At the highest level, the SSA family tree splits into two major
branches — those products with close ancestral ties to the
operating systems, software, and hardware architectures of
commodity servers and conventional hard disk arrays, and
the separately engineered, free‐standing enclosures or appli-
ances.

For example, you can take a commodity server, load it with an

operating system and other software designed or optimized
for flash, then fill its disk bays with SSDs — and you have a
solid state storage appliance. You can achieve essentially the
same effect by mating an enterprise storage array’s controller
with a disk enclosure filled with SSDs. These are both mem-
bers of the former branch of the SSA family tree.

The latter branch — purpose engineered boxes of flash

chips — are products that use less commodity hardware and
modified software, though many still use some. Most often,
a custom chassis is filled with units of flash storage — either
SSDs, PCIe cards, a memory module called a DIMM (dual in‐
line memory module), or custom‐designed modules. There are
versions of this design that also include hard disk drives to
lower costs and boost storage capacity.

No matter which branch of the family tree, SSAs aren’t

designed to be pushed into a server’s disk bay or plugged
into the PCIe bus. They stand alone. And that, frankly, is their
advantage. This is fast storage that is meant to be shared.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������ Chapter 2: Getting to Know Flash Storage Systems 19
They can be directly attached to a single server via one type
of interface or another, but more often SSAs connect to appli-
cation hosts within a SAN architecture.

Shared storage offers many benefits; shared fast storage offers

many more. Obviously, SSAs offer a much easier way than
SSDs or PCIe cards to provide low-latency, high-performance
storage to multiple servers — if you already have a storage
area network deployed. Though solutions are available, in
general it’s much more complex to configure ways to share
the performance and storage capacities of in‐server SSDs or
cards than it is to share storage from an SSA. Enterprises that
want to connect many servers, often groups or clusters of
such, and dozens or more applications to terabytes (TB) of
flash storage can do so much more easily with an SSA than if
they had to open every server and plug in PCIe cards or SSDs.

Because standalone flash arrays aren’t constrained by the

physical enclosure of a server or disk bay, nor by the hard
disk‐oriented interface protocols used in those cases, they
can contain a lot of very fast storage. Until recently, most
SSAs were rather dumb boxes, and the deployment options
almost always necessitated some connection to management
or controller devices. This wasn’t altogether a bad thing. It
was rather simple to hitch them to your SAN by making them
part of a storage array or by leveraging the storage manage-
ment functionality that has been evolving within host side
operating systems and related software.

The liability of having data streams travel through networks

and into and out of SANs may be overblown. SANs made
with Fibre Channel networks add only a few microseconds of
latency to the data’s round trips. In many cases, the design
of the software application itself, or any of a number of other
hardware and software components lying in the data path,
can add much more latency than the SAN. This fact probably
accounts for much of why SSA sales are so rapidly accelerat-
ing; even if connected in some networking fashion, they still
offer extremely low latency, and the simplicity of sharing their
resources is very attractive.

Plus, when you add storage through the network, solving data
protection and disaster recovery challenges becomes much
simpler. With your data storage in a separate pool, you can
easily copy it, then send the copies off to another machine,

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
20 Flash Array Deployment For Dummies, IBM Limited Edition _______

another building, or another city. When a hurricane sweeps in

off the Gulf and swamps your data center, your mission-critical
data isn’t lost. And finally, the SSA boxes are becoming a lot
smarter now days. This is the big new trend in all of enterprise
data storage — flash with all the bells and whistles.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Chapter 3

Choosing Flash Storage

Arrays
In This Chapter
▶▶Analyzing your storage needs
▶▶Performing tasks with flash controllers
▶▶Realizing the issues with SSDs and PCIe cards
▶▶Establishing the benefits of flash arrays
▶▶Introducing IBM FlashSystem arrays

T he decision to actually purchase and deploy flash ­storage

to support your enterprise has two parts. First, you must
assess your actual need. Then you must find the solution that
fits best. In this chapter, you first look at some of the storage
system analytic tools and resources available to help you
make the most accurate assessment possible of your stor-
age needs. Then you evaluate each of the flash‐based storage
options — SSDs, PCIe cards, and flash arrays — and find out
why flash arrays are strong candidates to address many enter-
prise data storage requirements.

Storage Analysis
After you see the need for quicker, more in‐depth decision
making, faster customer service, or a more palatable data
center budget, your next step is to accurately analyze your IT
infrastructure to identify exactly what kinds of system perfor-
mance issues you’re experiencing and where specifically they
lie. A key to lowering the risks and increasing the value of
your flash storage deployments is to thoroughly understand

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
22 Flash Array Deployment For Dummies, IBM Limited Edition _______

your system and application performance characteristics to

pin‐point where and how flash can offer the greatest value.

Most operating systems (OS) offer system monitoring and

diagnostic software programs or tools. Two of the most well‐
known over the years have been Performance Monitor (perf-
mon) for Windows and Iostat for the Unix family.

If you have a Unix‐flavored operating system such as Linux,

use the utility Iostat to perform analyses of your storage sys-
tem’s performance. Iostat is a computer system monitoring
tool within the Unix family used to collect and show operating
system storage input and output statistics. It is often used to
identify performance issues with storage devices, including
local disks or remote disks accessed over a network.

Specific applications, such as databases, also offer tools

that can help you better understand how your computer
systems, especially the storage devices, are operating and if
there are places within your hardware or software that are
creating problems or what are often called “bottlenecks” in
the performance of these systems. On the database side, the
most well‐known of these monitoring and diagnostic tools is
Statspack within the Oracle Database application. It is now
called Automatic Workload Repository (AWR). Reports gener-
ated by Oracle AWR provide database administrators (DBAs)
with detailed information concerning a snapshot of database
execution time. This snapshot furnishes statistics on wait
events, storage input and output volumes, and timings, as well
as various views of memory and activities associated with
software instructions to the database called SQL.

The statistics and insights provided by tools such as Oracle

AWR reports, as well as Iostat and perfmon, about the
memory, input and output (I/O), and SQL performance char-
acteristics are invaluable aids in determining if databases or
other applications and systems are functioning optimally.
From this type of information, you can make much more
informed decisions about if your IT environment can benefit
from adding flash storage, where specifically your perfor-
mance issues lie, and even some strong hints about what kind
of flash storage product might provide the greatest value.

But frankly, though your own systems administration staff

may be very familiar with and use these types of monitoring
and diagnostic tools, when you begin to seriously consider

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������������� Chapter 3: Choosing Flash Storage Arrays 23
deploying flash storage, you can take advantage of as much
help and expertise as you want — furnished by the flash stor-
age product vendors themselves.

All the legitimate flash storage solution providers in the

marketplace maintain technical experts often known as Sales
Engineers (SE) whose job it is to help you move successfully
along this path of information gathering, analysis, solution
design, testing, and deployment. Many of the larger product
vendors have made significant investments in these types of
resources. For example, just in the past few years IBM has
invested in laboratories around the world called Flash Centers
of Competency (CoC) where potential customers can get in‐
depth assessments that eliminate risks while maximizing the
deployment benefits of flash.

IBM Flash CoC teams offer comprehensive services to poten-

tial clients that involve detailed system and application work-
load analyses called Data Pattern Assessments. IBM experts
utilize data center analytics tools to execute end‐to‐end array,
host, database, and file scans of customer environments. IBM
Data Pattern Assessments require minimal investments of
time and resources from customers, but they return a rich
trove of information, which can greatly help you determine
valuable data points such as which applications, servers, and
storage volumes within your specific IT environment are most
impacted by storage performance bottlenecks and unaccept-
able latency and exactly how connecting your applications to
flash storage can provide the greatest benefit.

Flash Controllers
In order for flash to be viable as an enterprise‐grade storage
medium, several peculiar personality quirks of flash must be
mitigated and managed. This is the job of small processors
embedded within every flash storage product — the flash
­controller.

Flash controllers perform many tasks associated with writing

and reading data to the medium and managing various engi-
neering solutions that help make the particular flash product
faster, more reliable, and much longer lasting. Two of the

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
24 Flash Array Deployment For Dummies, IBM Limited Edition _______

most common flash management tasks are known as wear

­leveling and garbage collection:

✓✓Wear leveling in enterprise flash storage devices essen-

tially is the activity of spreading data evenly among flash
cells to increase flash life. A tremendous amount of very
innovative engineering has been focused over the past
decade or so on flash controller technologies in order to
optimize the useful life span or endurance of flash chips.
Unlike consumer uses for flash storage, such as in smart-
phones or digital cameras, enterprise use cases for flash
are characterized by a high number of program (write)
and erase (P/E) cycles. Flash would never work in enter-
prise environments if any particular cell was hit repeatedly
with new erases and writes. It would wear out much too
quickly. So enterprise flash product vendors design unique
wear leveling solutions into their flash controllers to
spread the P/E activity out over the hundreds of thousands
of flash cells in each device. Wear leveling has become so
effective that now enterprise flash storage wears out less
frequently than mechanical hard disk drives.
In a recent economic value validation performed on
IBM FlashSystem storage, the industry analyst firm ESG
estimated that mechanical disk drives wear out at a rate
of approximately 5 percent over a three-year period,
whereas flash modules in an equivalent IBM FlashSystem
array wear out at a 0.1 percent rate.
✓✓Garbage collection deals with the performance bottle-
neck that occurs due to the need for a flash cell to be
erased before it can be written to. To cut down on flash
cell write times and make flash as fast as possible, flash
controllers remember where invalid data exists, such as
data that has been updated elsewhere or deleted. Then
in the background, the flash controllers erase the cells
that contain invalid data and make them available for the
next writes coming into the device.

Wear leveling and garbage collection, among many other flash

management tasks, aren’t accomplished exactly the same
or equally well from storage vendor to storage vendor. The
speed, latency, consistency, predictability, reliability, and
efficiency metrics of each product provide some indication
of how well the flash controllers inside it perform. When you
load your mission‐critical data into the flash storage in which
you’ve just made a significant investment, these attributes will
grow very important to you.
These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������������� Chapter 3: Choosing Flash Storage Arrays 25

Looking at the Challenges of SSDs

After you accurately assess your data storage requirements —
how much storage capacity you need, how fast your data must
go to meet your business needs, where you have performance
bottlenecks within your system, and many other questions —
you can begin the process of evaluating your flash storage
options. I’ve already mentioned some advantages and benefits
of SSDs, but in this section, you look at some of their challenges.

Solid-state drives have been viewed as the most convenient

and lowest‐cost way to get flash into your system. But SSDs
do have their limitations and liabilities:

✓✓SSD form factors are very limiting. For example, wear

leveling across 10TBs works better than wear leveling
across 1TB because you have more flash cells between
which to spread out the writes. With a “box of flash” such
as a flash array, you can simply make the box bigger if
you need more flash capacity to achieve your objectives.
With a product designed to fit into the drive bay of a
server, this isn’t possible.
✓✓Because they are, in fact, intended to be deployed into
the same spaces as hard disk drives, they must use
the same interface protocols — essentially networking
languages — as hard disks, and these protocols usually
were not designed for the speed of flash. In general, all
the technologies built around disk drives work well with
latencies in the millisecond spectrum, from a few to
hundreds. Flash, on the other hand, operates in the
microsecond spectrum, ten to a thousand times faster.
Some components, either hardware or software, built
originally for disk speeds just can’t go at flash speeds.
So when you deploy flash in those environments, at least
some of the potential benefit of flash is wasted.
✓✓SSDs are rarely the whole solution by themselves. By
their nature, they’re intended to be plugged into some
larger device, whether a server or a storage array. This
larger system may not be optimized for flash, and so you
pay good money for bad performance.
✓✓The original advantage of SSDs was purchase price.
That’s still their advantage, but the enterprise market is
growing ever more knowledgeable about costs, and the
cost of storage involves more than the purchase price.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
26 Flash Array Deployment For Dummies, IBM Limited Edition _______

Operational expenses count too, such as the cost of elec-

tricity, the expense of cooling all those electronic devices,
the costs of software to manage the SSDs and the devices
that contain and manage the SSDs, and even the cost of
data center floor space. Other costs crop up. If you push
a bunch of SSDs into your servers, what are the costs of
implementing a solution to copy and protect the data on
all those individual SSDs? Because they don’t do wear lev-
eling as well, how much sooner will you be replacing them?
With SSDs you make cost trade‐offs — lower purchase
price versus lower overall costs — and you make per-
formance trade‐offs as well. SSDs aren’t optimized for
performance and because they only operate for you as
part of something else, a server disk bay or a SAN array
brings extra baggage in the forms of complex data paths
and added software that degrade the latency and overall
performance of the resulting storage solution.
✓✓Then there is the cost that everyone struggles to
define — the value of performance. SSDs, as a type of
storage device, are optimized for purchase price and
convenience, otherwise they wouldn’t use that shape
and those protocols. If you want the most performance
per dollar, you don’t deploy SSDs. In fact, if you want the
lowest price per TB of capacity, you don’t buy SSDs. As
every cowboy knows, ponies are cheaper than thorough-
breds, but you don’t take your pony to the horse race.

Understanding the Liabilities

of PCIe Cards
Just like SSDs, PCIe cards live by and at the same time suffer
from what in fact they claim to be — server‐centric applica-
tion acceleration. The concept hails from the days when one
application was hosted on one server. To make that one appli-
cation perform better, put faster storage in that one server.
Enterprises flocked to the idea.

But what about if you used multiple servers to host this appli-
cation? Server clusters, as these groups of computers are
sometimes called, became the first big engineering challenge
for PCIe cards. Yet at the same time, server virtualization was
gaining traction. This computer architecture involves loading

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������������� Chapter 3: Choosing Flash Storage Arrays 27
certain software onto the server to give it multiple personali-
ties. It pretends to be many computers instead of one, each
with its own OS and each able to host its own applications.
At first, PCIe cards thrived in virtualized environments, until
those environments included server clusters.

Of course, engineers soon provided various solutions to the

basic issue of sharing the data cached on PCIe cards segre-
gated in their individual machines. But one way or another,
the solutions involved networks, and now you had . . . net-
worked storage, which was exactly what the original concept
intended to avoid. Plus, implementing and managing all this
sharing between PCIe cards in different physical machines
involved lots of software, which to the extent that it intruded
into data paths and distracted CPUs thwarted the original
ultra‐low latency objective.

Add to these complications the added labor of needing to pry

open each individual server enclosure to install and maintain
or replace each PCIe card. Then mix in the limited physi-
cal space available within many of these server enclosures,
restricting the size, capacity, and capability of individual
cards, and enterprises have begun to seriously ponder PCIe
cards’ ratio of value to complexity.

Establishing the Advantages

of Flash Arrays
At this point, you can see that the directions where IT is
headed aren’t necessarily advantageous for SSDs or PCIe
cards. SSA sales themselves and their market share of over-
all flash sales tend to bear this out. But there are two larger
trends that will swamp these smaller differences between
storage devices.

✓✓First, everything related to IT is growing, expanding,

accelerating ever more rapidly. More data volume, higher
data velocities, more applications, more types of work-
loads such as mobile and social systems of engagement,
more of everything IT.
✓✓Secondly, software is growing smarter and this is enabling
the virtualization of all IT components. Essentially, this
is what Cloud computing means — CPUs are becoming

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
28 Flash Array Deployment For Dummies, IBM Limited Edition _______

a resource, networking is a resource, and yes, storage

is also becoming a resource to be managed, consumed,
migrated, updated, scaled up and scaled out, and
even subscribed to — all separately from the other
components.

How do you transform storage into such a resource with the

least amount of headache, complexity, and cost while at the
same time optimizing what you want from it — capacity and
performance? Enter the flash storage array.

Why you would choose to deploy flash storage arrays instead

of SSDs or PCIe cards includes many reasons:

✓✓Cost: The number one topic on everyone’s mind is cost.

SSDs offer the lowest purchase price; flash arrays offer
the lowest total cost, when both purchase price and
operational expenses are considered. Plus, flash arrays
offer the lowest price per capacity, if for no other reason
than there is less packaging per TB. You must buy many
separate SSDs to get 50TB of flash storage. You buy one
flash array a little bigger than a pizza box. Plus, you must
plug all those SSDs into something — something bigger
and more expensive than them, and almost certainly
not faster.
✓✓No slaves to multiple masters: If you’re going to imple-
ment networked storage, why not implement the least
costly and complex solution possible? Flash arrays are
simply optimized to share capacity and all their other
attributes with any and all applications that interface
with them. If you add more servers, fine. If you add more
storage, or reconfigure it, who cares? Flash arrays don’t.
✓✓Latency: If you want to accelerate the applications
hosted on one physical server, PCIe cards offer lower
purchase prices than flash arrays and better perfor-
mance than SSDs. If you already have or plan to imple-
ment shared storage, then you can’t beat flash arrays.

IBM FlashSystem
Before you make your decision about which flash storage solu-
tion to deploy, I have one more bone to pick. In the previous
chapter, flash arrays were introduced as descending from a

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ������������������� Chapter 3: Choosing Flash Storage Arrays 29
primal ancestor along two family tree branches — flash arrays
with many components, especially their hardware, not purpose‐
built explicitly for the role of flash array, and flash arrays with
all their components, essentially, purpose‐engineered for this
one role.

Why the difference? Cost, of course — the cost of develop-

ment and the cost of deployment. The truth is, most of the
flash arrays on the market today began as software engineer-
ing projects. A few ingenious software engineers built new
software that solved a particular storage problem. Then they
loaded it on hardware they essentially bought at the store.
If everything worked as planned, they had a competitive new
flash array product to sell, the result of relatively low‐cost
development coupled with quick time to market. Good
­business. But, good for you? Not necessarily.

If you want to combine the cost of flash and disk in your

storage deployment evaluation, simply buy the best of both
and integrate them with readily available automated stor-
age tiering software. You probably already have a disk‐based
storage array; just add more of the lowest cost disk you can
find. Then deploy an IBM FlashSystem array. This solution
offers lower cost per TB and much higher performance. It also
provides excellent storage virtualization and management
software, including dynamic tiering that automatically moves
data between storage media based on the policies set by you.
Figure 3-1 is a photo of an IBM FlashSystem V9000 with all the
storage bells and whistles included.

Figure 3-1: IBM FlashSystem V9000.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
30 Flash Array Deployment For Dummies, IBM Limited Edition _______

With the disks in their favorite environment and the flash

highly optimized within its IBM FlashSystem chassis, you can
add, change, and evolve this storage solution to meet your
evolving application and business needs without affecting any
other component of your IT infrastructure, including other
storage devices, and without ever throwing any disk or array
component away until they fail.

IBM FlashSystem all‐flash storage technology is purpose‐

engineered from the circuit to the chassis for the future of
information technology and of business itself. The hardware
descends from solid state storage ancestors going back liter-
ally decades. This is eons in IT time. The software in the base
IBM FlashSystem models can make the same claim. For the
model that replaces traditional enterprise storage arrays, IBM
has tightly integrated IBM’s industry‐leading storage services
and virtualization software into the IBM FlashSystem mix.
This software comes from a suite that has been deployed
­successfully in thousands of demanding IT environments
over the past decade. For example, the IBM Real‐time
Compression function alone is based on over 70 patents.
Then, IBM research and development labs around the world
are continually working to improve and enhance both IBM
FlashSystem hardware and software.

If you want to deploy flash storage at the lowest $/TB, look

long and hard at IBM FlashSystem. If you measure cost by
$/performance, then beating IBM FlashSystem won’t be easy.
Lowest possible latency? Check. Need to start small and then
grow your flash investment as your budget grows? Check.
Don’t have the manpower and need a solution that’s espe-
cially easy to deploy? Check.

Most importantly, IBM FlashSystem means that you can finally

move on from disk, and it’s cost effective. You can stop writ-
ing applications and architecting compute environments to
mitigate the shortcomings of traditional storage. You can
truly tackle the problem of endlessly rising power consump-
tion. You can fully exploit the potential of Cloud computing
and big data or implement virtual desktops without your
storage infrastructure getting in the way. In fact, at the end
of the day, that’s really the biggest IBM FlashSystem benefit:
Its performance, reliability, and efficiency turn storage from
a limiting factor into a real driver of innovation within your
business.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Chapter 4

Exploring Deployment
Designs with IBM
FlashSystem
In This Chapter
▶▶Learning how to directly attach IBM FlashSystem
▶▶Using SANs to your advantage

Y ou are an IT decision maker in a vibrant enterprise,

­perhaps an e‐commerce business, a hospital complex,
or an academic institution. You’ve familiarized yourself about
data storage technology and that journey led you to IBM
FlashSystem. Now, you must design and implement a success-
ful deployment strategy. Of course your business needs, and
how your current IT environment addresses and supports
them, guide you. You’ve already tapped into a worldwide
network of Flash Centers of Competency, Lab Services, Sales
Engineering teams, and solution architects who’ve analyzed
your needs, helped formulate the best solutions, and provided
resources and guidance for both off‐site and on‐site proof‐of‐
concept testing. To begin the process of developing the most
effective solution architecture, you and your IBM team have
to evaluate your various deployment options. This chapter
introduces the basic flash array deployment architectures
and provides some thoughts about why you may choose one
over the other.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
32 Flash Array Deployment For Dummies, IBM Limited Edition _______

Deploying Direct Attached

Storage with IBM FlashSystem
If you think that SANs are the only way that enterprises
deploy IBM FlashSystem arrays, think again. What if you’re
one of the hundreds of businesses around the globe that
directly engage in or provide IT support to online equities
trading? One application essentially defines your business.
Milliseconds define your timeframes. Moving at literally
lightning speeds and yet being able to guarantee the legally
required capture and ultra‐reliable storage of every transac-
tion are business requisites. Or you process data received
from a weather or research satellite? You analyze old seis-
mograph data looking for hidden petroleum reserves. You’re
installing smart meters for five million utility customers, and
this one application processes the multiple data streams
from each meter to better manage your portion of the power
grid. Direct attached flash storage may mean the difference
between discovering a new planet, increasing your revenue
by millions, or a blackout.

The direct attached storage (DAS) architecture is a venerable,

fairly simple storage solution design still used in many IT envi-
ronments, not only for business reasons but also for technical
reasons such as when connecting storage to large computers
called mainframes. Essentially, DAS refers to storage archi-
tectures where the storage device is linked directly to the
application host(s) with no or minimal intervening networking
resources. Normally, when you deploy a DAS storage solu-
tion, you don’t include a network switch, but instead cable the
storage device directly to the application server. You can see
an example of what DAS looks like in Figure 4-1.

Directly attaching IBM FlashSystem arrays is pretty simple.

Just follow a few steps:

1. Install Host Bus Adapters (HBA) in each server

where you want to directly attach IBM FlashSystem.
HBAs are hardware components with some software
that enable servers to interface with networks or send
signals directly through appropriate cables to another
device.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
  Chapter 4: Exploring Deployment Designs with IBM FlashSystem 33
2. Connect the HBAs to the IBM FlashSystem ports
(ensuring failover across all potential components)
perhaps with Fibre Channel cables or Ethernet
and carve capacity from the flash to present to the
server.
If the array is serving a single application or server
cluster, you can utilize the Open Access model of IBM
FlashSystem with enhanced management features to
simplify creating logical storage volumes called LUNs.
Access to IBM FlashSystem LUN provisioning and
­configuration is automatically open to all connected
servers.
3. Install the appropriate multipath configuration on
each connected server’s OS and utilize the new
­volumes as if they were any traditional disk.

Figure 4-1: The simple DAS architecture.

Directly attaching IBM FlashSystem storage has many

­benefits:

✓✓Lowest possible latency: HBAs directly cabled to other

appropriate devices generally add only 10 to 20 micro-
seconds (ms), which is only 10 to 15 percent of the
latency of the IBM FlashSystem unit itself.
✓✓Greater control: This is dedicated storage; you don’t
have to share it with anyone. Its performance and
­capacity are all yours if you own the application.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
34 Flash Array Deployment For Dummies, IBM Limited Edition _______

✓✓Higher capacity: You can enjoy memory‐fast storage

speeds without the physical capacity limitations of in‐
server solid state storage.
✓✓Greater reliability: Standalone arrays like IBM FlashSystem
offer better data protection, resiliency, and serviceability
than in‐server cards or SSDs.
✓✓Better failover: Server clusters and the software that
­manages this configuration offer the crucial “fail safe”
advantage of supporting “hot” backup virtual servers
or virtual machines (VM). If a physical server fails, all
its VMs can instantly migrate to their backups on other
physical servers. Your business critical applications
should never know that a server failed. But if a physi-
cal server fails, so do all of its in‐server storage devices.
Direct attached IBM FlashSystem has Active/Active ports
and controllers for each LUN, so as your VMs fail over,
storage remains available and unaffected — just like your
business.
✓✓Easier upgrades: With storage separated from servers
but not configured behind a Storage Area Network (SAN)
switch, you can add, subtract, and upgrade any IT com-
ponent without affecting or needing to upgrade others.
For example, you could upgrade your Fibre Channel
cabling to increase the bandwidth to your storage and
you wouldn’t need to install new networking switches or
other IT infrastructure.
✓✓Greater data security: Less infrastructure complexity
means less risk of failures that can affect your business.
✓✓Lower infrastructure costs: Just like with security, less
infrastructure complexity leads to lower costs.

It’s also important to point out the liabilities of DAS

architectures:

✓✓Less flexibility: Depending on how all your applications

are hosted, it may prove more difficult to share direct‐
attached IBM FlashSystem storage with all of them.
✓✓Lower utilization: It’s quite possible that IBM FlashSystem
performance exceeds that of the servers to which you’ve
connected it, leaving some of its capabilities untapped.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
  Chapter 4: Exploring Deployment Designs with IBM FlashSystem 35

Connecting Systems with a SAN-

Attached Architecture
The vast majority of enterprise IT infrastructures these
days are moving toward network attached storage, if they
already hadn’t years previously. Why? For many good
reasons, you don’t want a server; you want application
­hosting/data processing resources where, when, and how
you want them. The same holds true for enterprise storage.
Applications and their owners and users don’t need to know
about or care how you do it; they just want to get at their data
when, where, how, and how fast they need it. To enable this
revolutionary functionality, storage must be connected to the
other IT components through networking.

A powerful aspect of SAN-attached architectures is that you

can connect most any kind of appropriate storage system.
You will have your IBM FlashSystem array of course. You can
also connect hard disk drive arrays, often called RAID systems,
and even devices that write data to CDs or tape, usually for
backup or archive purposes.

To deploy IBM FlashSystem storage, you just install the

array(s) in a rack in your data center and connect it with the
proper cables (again Fibre Channel is the most popular) to
the SAN’s management software, often referred to as the name
server. The management software in the IBM FlashSystem
array introduces itself and makes the interfaces or ports vis-
ible. Most often, you then implement a feature called zoning in
the Active/Active mode, and IBM FlashSystem storage, perfor-
mance, and features will be available to your applications.

Your IBM FlashSystem storage becomes a resource, avail-

able to any device connected to the opposite side of that
SAN switch. But there are many other advantages of the SAN
deployment architecture:

✓✓More options: SAN deployment is the most common

model, which means there are more software and hard-
ware products available to support and extend it. For
example, almost all major switch brands and models are
qualified to support flash storage deployed in a SAN.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
36 Flash Array Deployment For Dummies, IBM Limited Edition _______

✓✓Easier expansion: A SAN enables more flexibility to

expand storage capacity and/or performance because
the same server ports can connect to multiple storage
systems.
✓✓Greater share‐ability: Any application/VM/server con-
nected to the SAN can leverage a portion of the IBM
FlashSystem resources.
✓✓Larger clusters and data sets: More and/or larger server
clusters can be implemented and data sets of all sizes
can be shared between multiple applications, server
clusters, or replicated around the world.
✓✓Easier scaling: There’s essentially unlimited room to
grow and tailor storage resources to match application
needs.
✓✓Higher throughput: Server clusters can use the full
throughput capabilities of the SAN when needed for
failover scenarios or spikes in application data traffic.

SAN-attached flash storage does have a couple of disadvantages:

✓✓Slightly lower performance: Because of the networking

involved, some, though minimal, performance impacts
are inevitable relative to direct‐attached and in‐server
solutions. Often this cost in latency is easily offset by the
greater flexibility of capacity/performance scaling and
easier/simpler deployment and maintenance.
✓✓Higher switch performance requirements: When you
deploy fast storage, all the other components in the data
path must be optimized to handle the increased perfor-
mance levels or your investment in flash will have less
impact.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Chapter 5

Implementing the Future

with Virtualized Storage
In This Chapter
▶▶Explaining the advantages of storage virtualization
▶▶Virtualizing all your storage using IBM FlashSystem

V irtualization is the future of enterprise data storage.

You focus on your business; your storage virtualization
engine focuses on increasing the efficiency, performance,
security, and accessibility of your data, while lowering its
cost. Storage virtualization offers some serious insulation
against the future. New storage media technologies have
already galloped over the far horizon of possibility; they just
haven’t climbed the nearer hills of cost‐effectiveness yet.
But if your storage is virtualized, and you view, manage, and
consume storage as a resource, no matter what the particular
storage medium happens to be, your virtualization engine will
manage it appropriately while your applications, and most
importantly your business, will never know the details.

IBM FlashSystem offers storage virtualization deeply inte-

grated with ultra‐fast flash — dozens of TBs of capacity all
in an enclosure the size of a few pizza boxes managed with
ever-increasing intelligence and sophistication, traveling
at the speed of integrated circuitry. You can purchase IBM
FlashSystem models that include the technology necessary
to virtualize all or any parts of the storage systems that com-
prise your SAN. After you deploy the array(s) and its ports
are presented to the name server, look to the easy‐to‐use IBM
FlashSystem graphical user interface (GUI) to virtualize your
existing storage under one management pane of glass, so to
speak. With this capability, you can then extend the ­powerful

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
38 Flash Array Deployment For Dummies, IBM Limited Edition _______

suite of IBM storage management features to all the other

storage systems in your SAN.

Virtualizing your storage using IBM FlashSystem changes your

SAN from a group of storage systems into an IT infrastruc-
ture resource that can be allocated, reallocated, scaled up
or down, upgraded, and on and on with no impact to and not
even any awareness of such by your many applications.

Storage virtualization offers benefits in a number of areas at

the heart of enterprise storage, including pooling, tiering, data
protection, and data reduction/capacity optimization:

✓✓Pooling brings storage resources together so that the

appropriate capacity can be delivered to each applica-
tion, and the magic of reallocating these resources is
enabled.
✓✓Tiering brings storage resources together so that the
appropriate performance can be delivered to each
application.
✓✓Data protection involves the multiple ways that enter-
prises ensure against data loss or corruption.
✓✓Capacity optimization most often utilizes “thin provi-
sioning” plus various data capacity reduction technolo-
gies to reduce the amount of idle or redundant data
stored and managed by your storage system, saving you
money in several ways.

Storage Pooling and Tiering

Essentially, storage virtualization enables automatic matching
of application workloads to the right storage resource. Before
storage virtualization, the data used by a particular applica-
tion, called its data set, was stored on a specific physical col-
lection of hard disk drives. To move that data set to another
storage resource, the application had to be turned off, then all
the information was moved, or migrated, the application was
updated with the new physical locations for its data, and then
everything was turned back on. The same might happen if
you wanted simply to add more storage capacity, say because
your application was growing — shut things down, reconfig-
ure, spin back up. It was very expensive, in terms of time and

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 5: Implementing the Future with Virtualized Storage 39
labor, but also because of the business or operations produc-
tivity lost while the application was offline.

Virtualization based on disk storage has enabled automatic

data migrations. But because fetching data from disks and
writing it to other disks is slow (certainly compared to other
IT components), these data migrations are more like the
movements of elephants or buffalo across the Serengeti Plain.
Virtualization enables them to happen without direct impact
to applications, but they don’t happen fast.

Hitching flash to storage virtualization changes things dra-

matically. Flash transforms data migration into data mobility.
Leading edge flash‐based storage virtualization such as that
integrated into IBM FlashSystem can move entire data sets but
also just portions of data sets, volumes, and sub‐volumes from
one storage resource to another very quickly. Now it’s the
gazelles darting and leaping, instead of the elephants plodding.

For example, some data sets become active only at certain

times — think of month‐end accounting applications. And
perhaps only certain portions of them. With virtualized flash,
to use the phrase that’s becoming fashionable, the virtualiza-
tion engine is constantly monitoring data activity and when
the end of the month rolls around and parts of that data set
become active, they can be moved quickly, automatically,
transparently from storage optimized for capacity, such as
disk or tape, to storage optimized for speed — flash.

A SAN can include multiple storage media. This is the way

SANs normally evolve. For discussion purposes, I’m going
to assume you started with a SAN composed of a single disk
storage array. Over time, you added another, or others, as
your business grew and/or diversified. Finally, performance
and cost factors caused you to add a flash array to the mix.
If it was the appropriate IBM FlashSystem model, then you
could bring all your separate storage systems together so that
to your applications they appear as a single pool of storage
resource. Within this pool of storage resource, your disk sys-
tems will be slow but relatively inexpensive per unit of bulk
capacity, and tape‐based systems will be even more so. Your
flash will be less expensive per unit of performance, much
less. And now you see the rationale for what is called tier-
ing. To lower costs and increase performance and efficiency,
you place data on the most appropriate storage medium.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
40 Flash Array Deployment For Dummies, IBM Limited Edition _______

Virtualized ­storage does this automatically, continually

seeking to maximize the utilization of your various storage
resources, whatever they may be, based on whatever poli-
cies you set as its priorities. Virtualized flash does this with
greater agility, leading to even greater savings and efficiency.

Data Protection
Another crucial function that your storage solution must per-
form in some manner is data protection. Essentially this means
that when your applications request it, your data is available,
and if some component or process fails within your IT infra-
structure, none of your data is lost forever.

Data protection actually becomes a very expensive proposi-

tion and is usually approached in two ways — preventing fail-
ures from happening, or at least from affecting the integrity of
your data, and making copies of the data so that the copy can
be used if the original is lost or corrupted.

To address the former and prevent failures, most enterprises

operate by the simple rule — no single point of failure. This
means that within the data pathways themselves, if any one
component fails, data will not be lost or corrupted. Because
hard experience has taught us that nothing is perfect, the only
way to ensure that no failure will result in lost data is to make
everything redundant. But a minimum of two of everything
drives up costs dramatically.

IBM FlashSystem helps you lower data protection costs by

engineering the arrays themselves with no internal single point
of failure. So, you make multiple redundant connections from
the SAN switch to the machine itself and data travels through
redundant pathways from the interfaces into the separate,
redundant flash storage modules. Even with this level of reli-
ability built in, some enterprises will still configure their stor-
age architectures with whole redundant systems. This type
of configuration is often referred to as mirroring, or deploying
a “hot spare” that can take over if the active system fails. But
IBM FlashSystem’s no single point of failure does offer the
option to forego the need to mirror or configure spares, and
this can result in much lower equipment purchase expenses.
As a matter of fact, achieving the no‐single‐point‐of‐failure
internal array architecture required years of engineering to

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 5: Implementing the Future with Virtualized Storage 41
accomplish and isn’t necessarily available on all other flash
storage arrays.

Internal hardware redundancy is not the only way that IBM

FlashSystem protects your data. The systems also employ RAID‐
based data protection regimes. Using this technology, a unit of
data is split into several parts and each is written to a separate
flash chip within a flash module of the array. Then a key, known
as a parity bit, is calculated by the controller and added to the
data unit. The parity bit enables reconstruction of the entire
data unit, if a flash chip fails and that part of the data is lost.

IBM FlashSystem uses a unique solution called Variable Stripe

RAID in each individual flash module. This innovation allows
the RAID algorithm to evolve if a chip fails, so that other flash
resources in the RAID group aren’t unnecessarily thrown out
with the failed chip, dramatically increasing efficiency and
lowering costs.

Then IBM FlashSystem goes another level better; it uses

RAID again, only at a system level between all flash modules,
instead of just inside each individual module. This means an
entire flash module could fail and you wouldn’t lose any data.
The two data protection components — module‐level Variable
Stripe RAID and system‐level hardware RAID — operate inde-
pendently, but together they provide synergistic system fault
tolerance to mend multiple flash memory failures.

No single point of failure, redundant components and data

paths, two dimensions of RAID, and these aren’t all the ways
IBM FlashSystem protects your data. Individual flash cells
aren’t all perfect; some hold a charge well and can be accu-
rately read, and some don’t. From the beginning of the use of
flash in mission‐critical environments, flash engineers com-
pensated for the lack of flash perfection with what is known
as Error Correction Codes (ECC). ECC algorithms are applied
by the flash controllers while data is being read to check for
errors and correct them on the fly. IBM uses a proprietary
“hard‐decision” algorithm to deliver very high correction
strength with lower processing overhead. The overall result
is that IBM’s unique error correction solutions drive up per-
formance, reliability, and throughput while driving down
­complexity and cost.

Even though IBM FlashSystem provides many layers of data

protection within the array itself, in general, storage systems

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
42 Flash Array Deployment For Dummies, IBM Limited Edition _______

haven’t been and many still aren’t so resilient. So enterprises

over the years have devised means outside of or not depen-
dent on the particular hardware or system to protect valuable
data. The most common of these is to make a copy of it and
store that copy somewhere else.

The two most popular ways to copy data sets are called
­snapshots and clones. Snapshots involve essentially taking
quick pictures of the data set at specified moments in time.
Then, if data is corrupted, the system can be moved back in
time to the last snapshot and started again with data that was
correct at that point. Obviously, the more often you do snap-
shots, the more recent your backup will be. But of course you
must store these snapshots, which takes resources away from
your primary application workloads.

Snapshots lead to two challenges that storage virtualization

addresses especially well. First, the process used to reduce
the storage resources needed for snapshots can cause signifi-
cant impacts to storage performance because they involve
more processing and software in the data path. Storage vir-
tualization can dynamically move snapshot activities out of
“the line of fire” so to speak, utilizing storage resources avail-
able at any particular moment that will least affect latency.
Virtualized flash goes one better — it offers extra performance
and lower latency so that snapshots can be done in flash‐
based resources with minimal impact on overall performance.

Next, think about the situation in most SAN environments —

they have multiple systems, some flash, some disk or tape,
each different, and often none that “talk” to each other. How
can we manually perform a coherent snapshot across all
these disparate systems? With storage virtualization we can,
because the storage is managed as one resource, not as dif-
ferent systems. Virtualization tools such as IBM FlashCopy
Manager can synchronize and manage snapshots across
disparate slower and faster arrays and use IBM FlashSystem
resources to almost eliminate performance impacts.

Clones are another data protection strategy that storage

virtualization enables. Clones are complete copies of the
entire data set, very different from space‐efficient snapshots
that may just capture the changes to data. Clones are used
to recover from disasters and major system failures. Another
important use is for the software development, testing, and
new application qualification environments, which are c ­ arefully

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ���� Chapter 5: Implementing the Future with Virtualized Storage 43
segregated from the actual production environments but still
need to use a relatively accurate or legitimate version of the
data set. Storage virtualization enables data set clones to be
“shipped” to separate storage resources whenever needed for
software development and testing with no impact on the pro-
duction environment. Appropriate storage resources can be
quickly allocated and configured for these use cases based on
the capacity and performance needed and/or available.

Capacity Optimization
Historically, storage capacity has been a static resource. You
have this much, period. To add more, you must stop every-
thing, physically haul in and configure more disks or new
systems, then spin it all up again and hope nothing explodes.
To avoid the risk of running out of storage capacity unexpect-
edly and to account for growth, you allocate or provision
a lot more than you actually need right now. This is called
“over‐provisioning,” a venerable and expensive storage man-
agement practice that can result in a lot of resources spinning
happily away unused.

Thin provisioning means allocating only the storage resources

you need right now, the opposite of over‐provisioning. It’s
much more efficient and less expensive, but in traditional
storage environments it’s too risky. Not when you’ve imple-
mented storage virtualization. When you can add capacity
quickly and easily, the storage world flips.

With virtualized flash such as IBM FlashSystem, when you

deploy and configure this new technology and make all of
your storage a single resource, you literally try to find ways
to allocate 100 percent of the IBM FlashSystem capacity. You
want all of that extraordinary performance working for you,
right now. IBM FlashSystem comes with thin provisioning
technology that, in fact, allows you to over‐allocate its capac-
ity. Thin provisioning functionality carefully monitors actual
storage usage and automatically allocates more from other
LUNs or other available systems just when needed, then
allocates it elsewhere when no longer needed. If a call comes
in at 2:00 a.m. on Saturday that data volumes are reaching
90 ­percent utilization and climbing quickly, external storage
can be easily zoned in by storage virtualization and utilized for
data growth without panic or even a trip to the data center.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
44 Flash Array Deployment For Dummies, IBM Limited Edition _______

Another storage capacity reduction technique is called

data compression. Compression is the reduction in size of
data in order to save space or network transmission time.
Applications write data to storage and during the write pro-
cess a tool such as IBM Real‐time Compression shrinks the
amount of storage capacity needed by implementing software-
or hardware‐based formulas that remove all extra-space
characters, insert a single-repeat character to indicate a string
of repeated characters, and/or substitute smaller bit strings
for frequently occurring characters, among many other tech-
niques. IBM Real‐time Compression can reduce certain types
of data files by a ratio of up to 5:1.

Because flash is still more expensive per unit of capacity than

some disk storage, data compression and other reduction
strategies tend to offer even greater benefits when applied to
flash storage. When data compression is implemented using
software running on commodity processors, it can signifi-
cantly impact the storage latency. IBM FlashSystem there-
fore implements IBM Real‐time Compression using a mostly
hardware‐based process, which minimizes the latency impact
while maximizing the degree of compression. Also, some data
types don’t yield much benefit from compression algorithms.
During deployment, or at any time afterward, the virtualiza-
tion engine within IBM FlashSystem allows you to enable IBM
Real‐time Compression only on the data volumes you specify,
thus optimizing their performance.

The flexibility of data compression deployment offered by IBM

FlashSystem results from a large group of IBM innovations col-
lectively referred to as IBM FlashCore technology. These vari-
ous innovations enable IBM FlashSystem to deliver the wide
range of operational and cost efficiencies, such as the agility
of Real‐time Compression. IBM FlashCore technology lies at
the heart of FlashSystem storage. Fundamental to this technol-
ogy is the concept of the hardware‐accelerated data stream
that delivers very high performance while also supporting the
capacity optimization features essential to modern enterprise‐
class storage.

Because the engineering embodied in IBM FlashCore tech-

nology is so strong and yet so flexible, it enables IBM
FlashSystem to incorporate new performance and capacity
optimization features, as well as many other virtualization
capabilities, with absolutely no compromise in system perfor-
mance or reliability for many years into the future.

These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
These materials are © 2015 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.