White Paper

Architectural Design Decisions Directly Support a Better

Customer Experience for Pure Storage FlashArray Users
Sponsored by: Pure Storage
Eric Burgener
September 2020

IDC OPINION

In all-flash arrays (AFAs), architectural design choices can have a major impact on the quality of the
enterprise storage life-cycle experience. Pure Storage, a leading vendor in terms of enterprise-class
AFA revenue, changed the industry forever with the introduction of its Evergreen Storage program
back in 2015. The program gives customers a number of guarantees and investment protection
offerings that enterprise storage providers had not provided to their customers before. These elements
range from guaranteed data reduction ratios and overall customer satisfaction to nondisruptive
upgrades and multigenerational technology refresh, capacity consolidation, included hardware
upgrades, and on-demand trade-in credits. Evergreen Storage has prompted responses from Pure
Storage's competitors, which are now offering programs of their own. In Pure Storage's case, many of
the guarantees and offerings are not just marketing programs, they are directly supported by specific
design choices and features in the company's FlashArray system architecture.

This white paper takes a closer look at FlashArray's architectural support for the Evergreen Storage
program, delving into four very specific areas: software-driven architecture, stateless controllers,
modularity of components, and simplicity of design. Real customer benefits — whether in terms of
better performance, higher availability, easier deployment and management, improved efficiencies,
nondisruptive multigenerational technology refresh, and/or lower costs — result from these architectural
choices. These benefits are also a key factor in Pure Storage's industry-leading Net Promoter Score,
an objective third-party metric that measures the quality of the overall customer experience (CX).

The typical legacy enterprise storage life cycle for a SAN array is 4–5 years, after which a need to
move to newer technologies commonly drives an often painful "forklift" upgrade. Pure Storage claims
to have doubled the enterprise storage life cycle for its FlashArray to 8–10 years, simplifying the
enterprise storage life-cycle experience and significantly lowering the total cost of ownership. The
flexibility to accommodate newer technologies nondisruptively is a key determinant of array longevity,
and Pure Storage has a strong case to make for its FlashArray that is directly supported by
architectural design decisions. As enterprises consider new storage platforms, they should seek a
modernized version of the enterprise storage life cycle and expect vendors to validate their ability to
deliver that.

September 2020, IDC #US46800220

IN THIS WHITE PAPER

Architectural decisions can make a huge difference in providing the performance, availability, efficiency,
upgradability, and simplicity that effectively differentiate a storage system from those of competitors. In
the case of Pure Storage's FlashArray, a number of architectural decisions were made that have allowed
the storage platform to deliver a customer experience that is clearly different from those offered by legacy
enterprise storage platforms. This white paper summarizes those architectural decisions, discussing
implementation as well as the resulting impacts on CX. In addition, the paper explains the reasons for the
architectural choice and summarizes the relevant impact on performance, availability, efficiency,
upgradability, and/or cost. This discussion is specific to Pure Storage's low-latency block-based
FlashArray storage platform. Readers should keep in mind that there are five FlashArray models
currently being sold: //X10, //X20, //X50, //X70, and //X90. Each model uses the same architecture but
differs in the power (the performance) of its controllers and the amount of raw storage capacity a system
can support. All FlashArray systems are end-to-end NVMe, with optional NVMe over Fabrics host
connectivity or SCSI-based host connections (using Fibre Channel or Ethernet).

SITUATION OVERVIEW

Information technology (IT) organizations that manage their own enterprise storage systems are very
familiar with the legacy enterprise storage experience and life cycle. Over the years, storage managers
have almost become inured to working with less flexible hardware-defined architectures, paying
separately for various data services (snapshot integration, replication options, etc.) and managing
systems through "nerd knobs" that require significant storage expertise. They have also endured
maintenance price increases as a system ages, and then roughly every four or five years they are
forced to go through a time-consuming and disruptive forklift upgrade process to move to newer
technologies. Although not all of these aspects of enterprise storage ownership flow directly from
system architecture decisions, many of them do. In a word, architecture matters.

When the first enterprise-class AFAs began to ship in the 2011–2012 time frame, the use of solid state
storage as a persistent storage medium set these systems apart from their legacy brethren that were
still architected for (and used) hard disk drives (HDDs). Some storage vendors just inserted NAND
flash–based solid state disks (SSDs) into systems that had originally been designed around HDDs.
Others made more sweeping architectural changes that made better, more efficient use of the new
solid state storage medium. As one of the AFA start-ups that architected its systems from a "blank
sheet of paper," Pure Storage made a number of design decisions intended to improve overall
performance, availability, efficiency (of resource utilization), and simplicity that were directly linked to
the decision to use only persistent flash media. One of the business objectives of Pure Storage's
design intent was to at least double the useful life of an enterprise storage system, further enhancing
the cost benefits associated with its more "flash optimized" design.

In considering Pure Storage's architectural decisions that have a direct impact on extending the
storage life cycle, IDC notes that the vendor created and marketed an entire ownership program,
called Evergreen Storage, to underline the advantages its design offers across a number of areas. The
name "Evergreen" was chosen because the offering did at least double the useful life of a storage
system while still allowing that system to nondisruptively accommodate newer storage technologies as
they became available — in other words, the system stayed "evergreen" throughout its entire life cycle.
Many of Pure Storage's competitors have characterized Evergreen Storage as "just a program," but a
deeper look at the vendor's system architecture disputes that claim.

©2020 IDC #US46800220 2

Software-Driven Architecture
Software-driven infrastructure has become a popular topic among vendors, particularly software-
defined storage (SDS) vendors, in recent years. Compared with older, more hardware-defined
designs, software-defined designs put a storage system's functionality and features (beyond just
persistence) into software. With software-defined functionality, system capabilities in the areas of
performance, availability, capacity, and storage management can be upgraded via software. This
functionality makes it not only simpler but also less expensive to add new features to an existing
system in the field because it does not require the purchase of any new hardware. This does not mean
that system enhancements can never be tied to new hardware — for example, buying additional
storage devices adds capacity — but software-defined functionality provides a much greater flexibility to
upgrade system capabilities without requiring additional hardware. And even in the case of additional
capacity, software upgrades that improve data reduction ratios can add effective capacity to a system
without requiring any hardware purchases.

Limited support for different storage device geometries can be a problem with some legacy systems,
precluding them from supporting newer storage technologies that may deliver faster performance,
increased density, or other attractive capabilities. When storage device support is in some way limited
by the storage controllers, controller upgrades may be required to support newer device geometries.
Most systems do offer a "midlife kicker" that includes a "new" controller at some point in a system's life
cycle that may introduce incremental improvements, but these are often not able to accommodate
significant advancements like moving from SAS to NVMe or accommodating larger-capacity devices
and/or new media types. In some systems, the only way to obtain a controller that can support desired
new technologies is to perform a forklift system upgrade. When the software is the critical factor in
storage device support (rather than the controller), there is significantly more flexibility to
accommodate newer technologies.

Software-defined designs also enable more flexibility in handling metadata. If the metadata is
extensible, a system can accommodate on-disk format changes that improve the performance,
endurance, and/or efficiency of operation gains without requiring any data migration. If the metadata is
static and tied to the media format, then there is little opportunity to change it to improve system
capabilities without migrating data. More seasoned storage administrators may remember cases
where they were forced to migrate all of the data on an array to obtain needed new "software" features
because the on-disk format was changed with the new software release. Vendors are well aware of the
pain this causes, but in some cases, the need to extend the capabilities (usually the capacity
scalability) of the system is deemed to outweigh the pain.

The ability to take advantage of new technologies that offer significantly improved capabilities through
just a software upgrade is one aspect of extending the storage system life cycle, and Pure Storage has
an excellent track record here. Since 2012, new software releases of Purity FA, the FlashArray storage
operating system, have improved performance, increased scalability through (among other things)
support for new device geometries, accommodated new solid state media types (MLC to TLC to QLC),
increased data reduction ratios, and enabled key new features, like cloud-based predictive analytics
and support for native file systems, replication, and stretch clusters. The FlashArray's metadata is
extensible, and over the years, the company has continuously evolved it to make its systems more
performant, scalable, and efficient. All of these changes have been implemented in a manner that
allows the installed base of systems to upgrade nondisruptively, helping keep the systems
"evergreen."

©2020 IDC #US46800220 3

The FlashArray's ability to simultaneously support multiple device geometries offers what Pure Storage
refers to as "Capacity Consolidation." For an older FlashArray, a customer may at some point want to
move from lower-capacity to higher-capacity (i.e., more dense) storage devices. This is an evolution
that in competitive SAN systems often cannot be done and, even if it can be, often requires manual
data migration. On the FlashArray, a customer can insert a new disk shelf (assuming that space is
available) of the newer storage devices, and the system will, over time, migrate data from the older to
the newer devices as a background process so as not to impact application performance. Once the
data movement is complete, the administrator is notified, and the older shelf may be removed. In this
way, a system that takes up the same amount of rack space can be expanded over time not only to
higher-performance controllers (using the Evergreen Gold subscription's Upgrade Flex and Free Every
Three features) but also to increased storage density (using the Capacity Consolidation feature).

Stateless Controllers
Many legacy storage arrays use a stateful controller design, an architecture that complicates controller
replacements and upgrades. For example, if the memory that holds controller state information is held
in the controller itself, removing a controller severely impacts array performance or even the ability to
remain online. In the FlashArray, controller state information is kept in separate cache cards (which are
NVMe based and can use either NAND flash or 3D XPoint media, depending on customer
performance requirements and FlashArray model). When a controller is replaced, data on the NVMe-
based cache cards is very rapidly reloaded onto the controller and the array is moved back into normal
operation mode. This stateless controller design is a key element in enabling FlashArray to be
upgraded nondisruptively via controller swap out. The fact that state information is kept separate from
the controller also provides better data security than stateful controller designs.

FlashArray also uses a modified "active/passive" controller architecture where, during normal
operation, both controllers accept I/O from the initiator side. The two controllers coordinate their
operations such that data is moved to and from storage devices in a system through only a single
controller. If a controller fails, the remaining controller continues to service all I/O with no change in
performance. Storage administrators familiar with "active/active" controller designs may observe that
having both controllers active during normal operation drives higher performance, and this may be
true, depending on the power of the respective controllers being compared. Readers should note that
in building each class of storage controller, Pure Storage designed each single controller to be able to
handle the same amount of I/O as its competitors' dual controller designs (when operating normally).
Another way to state this is that a single FlashArray controller is roughly twice as powerful as Pure
Storage's competitors' offerings on a per-controller basis. The system was designed this way so that
Pure Storage could provide the benefits of the "active/passive" design (a failed controller causes no
performance impact) without the downside (a system runs only half as fast as during normal
operation).

FlashArray controllers can be upgraded at any time during the life of an array to move to more
powerful, latest-generation controllers within the same class or to move to a higher, more powerful
class of system, also of the latest generation. (The "upgrade at any time" option under Evergreen Gold
that includes full trade-in credit for old controllers is called Upgrade Flex. Evergreen Gold also includes
a feature for no-cost controller upgrades to the latest generation of the same class of controller, called
Free Every Three.) For more information and IDC commentary on the Evergreen Program, see Pure
Storage's Evergreen Storage Continues to Be the All-Flash Array Market's Customer Experience
Program to Beat (IDC #US45761920, January 2020).

©2020 IDC #US46800220 4

Pure Storage's approach to controller design in FlashArray has several positive implications relative to
"active/active" designs. First, upgrading a system to new controllers does not impact application
services. The FlashArray architecture is such that a low-end //X20 system can be nondisruptively field
upgraded to an //X50, an //X70, and even an //X90 without ever shutting the system down or impacting
application performance. There are scenarios in which customers require more performance, not to
mention higher capacity potential, from their arrays as workloads increase and requirements change
over the life of an array. If that need arises, customers have the option to nondisruptively upgrade
controllers to get to the next performance level (i.e., move from an //X20 to an //X50, from an //X50 to
an //X70, etc.) without a forklift upgrade or a data migration. Readers should note that this upgrade
capability also leverages the stateless controller design and modular components (discussed in the
section that follows), both of which are featured in FlashArray. Second, this capability has a positive
impact on the reliability of the controller pair. During normal operation, only one of the controllers is
communicating with the back-end storage devices, not both. Because systems are operating normally
instead of in failure mode most of the time, there is less overall stress on the controller pair (which
combined run at a lower level of utilization) than with competitive "active/active" controllers in the same
class (which are both handling initiator and target traffic during normal operation). Running
components at lower levels of utilization has a positive effect on reliability.

Modularity of Components
FlashArray employs a very modular design in which major components that affect system
performance, availability, and scalability can be hot-plug replaced quickly and easily, without tools.
Controllers, cache cards, storage devices, disk shelves, power supplies, and fans are redundant where
meaningful (the disk shelves aren't necessarily redundant) and provide the hardware foundation for the
system's "six nines" availability track record. While the FlashArray backplane cannot be field replaced,
starting in 2016, it could support both SAS and NVMe storage devices. After Pure Storage shipped its
first NVMe AFAs (NAFAs) in 2017, existing customers began upgrading their installed base systems
from SAS to NVMe, all completely nondisruptively. Although their competitors began shipping their first
NAFAs in 2018 and through 2019, none of them ever allowed their installed base of SAS systems to
be nondisruptively upgraded in place to NVMe — they all required a forklift upgrade to get to NVMe.
FlashArray's nondisruptive SAS-to-NVMe upgrade path was unique in the industry, and no other
vendor has offered this level of "evergreen" upgradability for in-place scale-up systems.

Systems that are more modular in design offer more opportunities to upgrade different storage
components to newer technologies over the life of an array. Take the vendor's cache cards as an
example. Like many enterprise storage vendors, Pure Storage uses solid state storage for its cache,
sending all writes there first to take advantage of the extremely fast write acknowledgements that the
medium can provide. Competitive caches in an array are often on the storage controllers, but Pure
Storage's use of separate cards allowed the company to upgrade cache cards without upgrading
controllers and work with higher-capacity caches that could handle all writes and also a good number
of reads. In early systems, when the rest of the array was plumbed with SAS, the cache cards used
NVMe to provide lower latencies. As the vendor shipped more NAFAs with larger capacities, Pure
Storage introduced an optional storage-class memory-based cache card that lowered cache latencies
yet again by an order of magnitude (available specifically on the //X70 and //X90 models). This new
cache card, called the DirectMemory Cache, became available in September 2019, and FlashArray
systems can support up to three of them. And in the spirit of its modular upgradeable design, previous-
generation //X70 and //X90 systems can be retrofitted with DirectMemory Cache.

©2020 IDC #US46800220 5

The name of the DirectMemory Cache plays off the name of Pure Storage's custom-built storage
devices, the DirectFlash Modules (DFMs). In 2017, Pure Storage moved away from off-the-shelf
SSDs, sourced from major storage component suppliers for the FlashArray, to the new custom DFMs.
All along, the FlashArray's storage operating system had used some proprietary algorithms to manage
the flash media directly for improved performance, higher endurance, and better capacity utilization.
But the company could not always get the access it wanted to the underlying media when working with
off-the-shelf SSDs. The SSD controller and flash translation layer provided by the storage component
suppliers were often "in the way." Moving to the custom DFMs gave the company an unfettered ability
to manage the underlying flash media directly (hence the "Direct" name), driving performance and
efficiency improvements, while the custom form factor allowed Pure Storage to introduce higher-
capacity devices for increased storage density than was generally available with off-the-shelf SSDs.
Today, the largest enterprise-class SSD in common use is around 15TB, while Pure Storage sells an
18TB DFM for the FlashArray//X (as well as DFMs as low as 1TB in capacity). Both the DirectMemory
Cache and the DirectFlash Modules are hot-pluggable.

The ability to manage the underlying solid state media directly in a media-agnostic manner is important for
several reasons. First, it allows Pure Storage to implement I/O algorithms that are best suited for efficient
use of the underlying media given the I/O patterns in its target markets and associated use cases. Pure
Storage is not forced to work through SSD-resident controllers and flash translation layers that tend to be
more attuned to the high-volume consumer rather than enterprise market requirements. This has
implications not only for latencies but also for how garbage collection (i.e., free space allocation) and media
endurance are managed — issues that are particularly important as Pure Storage moves to take advantage
of other types of NAND flash–based media (TLC and QLC) that have lower performance and endurance.
Second, by optimizing media endurance, the company can lower media overprovisioning ratios to cut costs
out of the storage devices without sacrificing device-level longevity (the vendor overprovisions media at a
lower rate than the established storage component suppliers but provides a lifetime-of-the-system
guarantee on all solid state devices it sells). Third, the ability to manage underlying solid state media allows
Pure Storage to accommodate new media types faster because it builds them into the DFMs and qualifies
them for production use faster than competitors that must wait for storage component vendors to build and
release a device before they can begin their own system-level qualifications.

Simplicity of Design
The FlashArray leverages a simpler design than the offerings of most of Pure Storage's competitors in
the large enterprise systems market. First, the company has stayed with a dual controller design, while
many of its competitors moved to a more complex design that allowed them to scale controllers
beyond just two. Through the use of either a crossbar switch or a highly specialized ASIC, competitive
systems could scale up to four and sometimes as many as eight controller pairs to extend the
performance of their systems in place. In Pure Storage's case, customers needing more performance
could just quickly and nondisruptively upgrade to a higher class of storage controllers. The crossbar
switch and/or ASIC designs added significant complexity and even reliability issues to the competitive
systems, both of which must be managed. Because of cache coherency overhead and other east-west
traffic, these multicontroller systems did not scale performance linearly. The FlashArray uses existing
system resources more efficiently with the much simpler dual controller design. While it is true that
Pure Storage may not be able to build a single system that can drive as many IOPS as some
competitive offerings at the very high end, many customers actually prefer to split such large
workloads between the fault domains of multiple systems for availability reasons. For those
organizations that prefer to consolidate tens of millions of IOPS onto a single system, Pure Storage
may not be the best fit.

©2020 IDC #US46800220 6

Traditionally, large enterprise storage systems have included administrative interfaces that offer very
deep and granular configuration flexibility. These "nerd knobs" (as they are often referred to by
sophisticated storage administrators) were originally needed because there was such a disparity
between the performance of compute and storage (i.e., HDDs) resources that customers needed to
continuously fiddle with settings to optimize performance as workloads evolved over time. When all-
flash systems were introduced, one of the first things that storage administrators noticed was that, with
the predictably consistent sub-millisecond latencies AFAs delivered under load, almost all storage
performance problems just went away. Vendors that introduced AFAs by upgrading their existing HDD-
based architectures kept these "nerd knobs" because they were already a part of their storage
operating systems, but many AFA start-ups took the opportunity to think differently about what they
demanded from storage administrators.

Pure Storage took this latter approach with its FlashArray. Instead of making everything configurable, the
vendor chose a set of defaults that allowed a system to be unboxed and deployed very rapidly. Most of
the enterprise-class data services, which in the past had to be individually configured, were now just "on"
all the time. With HDD-based systems, there was a noticeable latency impact with many data services
choices, but the performance of flash made that a nonissue at the application level for almost all
workloads in AFAs. For example, instead of having to decide which storage device–based data protection
method to use from a list, which included RAID 1 (mirroring), RAID 5 (single-parity RAID), RAID 6 (dual-
parity RAID), or RAID 10 (mirrored stripes), and actually implement that method, Pure Storage assumes
and implements a modified RAID 6 implementation for all data — automatically and without any operator
intervention required. Features that were performance sapping in HDD-based and legacy-retrofitted AFA
environments like thin provisioning and inline data reduction were instead written into the Purity OS to be
"always-on" without impacting application performance in FlashArray.

While this is a great enabler for less experienced storage administrators (which by the way are
becoming much more popular as the responsibilities for storage management in many IT shops are
migrating more toward IT generalists like virtual, Linux, and Windows administrators), what about
seasoned storage administrators who are comfortable with "nerd knobs"? Generally, administrators
who may express a preference for them up front do so because, based on experiences with HDD-
based arrays or legacy architecture AFAs, they do not trust that default configurations will give them all
the performance they need for particular workloads and thus may keep them from meeting certain
service-level agreements (SLAs).

Pure Storage's all-flash design effectively resolves this issue. Customer prospects concerned about
this issue can talk to their colleagues who are using FlashArrays (there are tens of thousands of
systems in the field). They will find that FlashArray systems can deliver sub-millisecond latencies at
scale (i.e., while pushing millions of IOPS), and if they still have concerns about storage latencies, they
can add optional DirectMemory Cache cards, which bring storage latencies down to the 150-
microsecond range. Ultimately, even sophisticated storage administrators come to appreciate the
simplicity of the FlashArray's default configurations once they see that they do not in most cases
impact their ability to meet storage SLAs. And having features like inline data reduction on at all times,
which can be guaranteed by Pure Storage and drives an average 5:1 data reduction ratio in mixed
workload environments, produces very desirable economics. These are all critical parts of the
FlashArray value proposition — that customers will not have to give up performance to get simplicity of
management, efficient use of storage resources, or a compelling total cost of ownership (due in part to
high data reduction ratios).

©2020 IDC #US46800220 7

On a side note, when evaluating the total cost of ownership of any system, from Pure Storage or any other
vendor, IT managers should focus on the effective capacity (the capacity available for use after formatting,
data protection, and inline data reduction have been applied), not the raw capacity. Raw capacity
advantages can quickly be overcome by higher data reduction ratios, particularly in larger configurations.

There are other simplicity aspects of the FlashArray CX that should be pointed out. Each independent
system is one physical box with six cables that can be installed in 30 minutes. All array software, even
advanced features like ActiveCluster (Pure Storage's stretched cluster option), are included with the
base price of the array. Features like thin provisioning, inline data reduction, quality of service, and
FIPS 140-2–compliant end-to-end encryption are always on. ActiveCluster uses a cloud-based
mediator to resolve split-brain syndrome, removing the need to buy and maintain a separate quorum
device for these configurations. Snapshots are "portable" in the sense that they are incremental, space
efficient, and self-describing and can be shared between other FlashArrays, FlashBlades (the vendor's
scale-out unstructured storage platform), third-party NFS storage, or the public cloud. Pure1, an AI/ML-
driven cloud-based predictive analytics platform providing full-stack analytics, is included with the base
price of each array. And FlashArray includes APIs that enable integration with VMware, OpenShift, and
other automation/orchestration tools.

The Impact of FlashArray Life-Cycle Extension

IDC has commented at length about the features and benefits of Pure Storage's Evergreen Storage
program under a separate cover. Aspects of that program, like all-inclusive software and guarantees
for data reduction ratios and fixed maintenance pricing for the life of the array, clearly simplify the
overall experience of owning and managing an enterprise storage array. Other features of that
program extend the life cycle of a FlashArray while still enabling customers to take advantage of new
technological advancements. The Evergreen Gold program provides an included controller upgrade
every three years under the Free Every Three option, and the Upgrade Flex option allows customers to
upgrade controllers at any time with full trade-in credit for old controllers. The Capacity Consolidation
option lets customers move to denser and more modern storage devices in place to expand system
capacity without increasing footprint. Unlike competitive programs, Pure Storage provides trade-in
credits for up to the full amount paid for the components being retired toward the purchase of the new
components, which means customers are not rebuying controllers and/or capacity at regular intervals
— investment in these components is truly preserved.

It is the combination of these features that double the storage life cycle of a FlashArray into the 8- to
10-year range. In competitive systems, forklift upgrades are required when existing technologies run
out of gas and can't be upgraded in place — a cycle that happens roughly every 4–5 years. Limitations
around backplanes (SAS only), controllers, and supported storage device geometries are all factors
that drive this legacy forklift upgrade, but on the FlashArray, these factors do not impede technology
refresh. A SAS-based system purchased in 2016 from Pure Storage that used off-the-shelf SSDs
could be nondisruptively upgraded to an end-to-end NVMe-based system that uses DFMs and
supports newer device geometries (for increased storage density) and different media types (from
MLC to TLC and/or 3D XPoint) — all without having to rebuy performance and capacity that are already
paid for. Performance and capacity upgrades do generate some incremental costs, but those are
always defrayed by the guaranteed trade-in credit amounts that customers know up front when they
buy their FlashArray. These are the features that support the vendor's claim that a FlashArray has an
8- to 10-year life cycle relative to conventional SAN arrays, enabling the nondisruptive integration of
the latest technologies to ward off the need to move to a completely new system.

©2020 IDC #US46800220 8

The extended life cycle has very positive impacts. Storage administrators know the pain involved in
data migration during technology refresh, and Pure Storage's approach means that, over the life cycle
of a single array, a relationship with the vendor would cut the need for that at least in half. Customers
are not rebuying entirely new storage systems and replacing storage capacity every 4–5 years. IDC
would encourage prospects that understand the technology behind the Evergreen Storage program to
compare the total cost of ownership and the management impact of this life-cycle extension to that of
the traditional 4- to 5-year enterprise storage life cycle.

CHALLENGES/OPPORTUNITIES

For more seasoned IT personnel, it may be easy to dismiss Pure Storage's claims about the value
driven by the Evergreen Storage program as "just another program." Because of the early popularity
with customers of Evergreen Storage, competitors across the board have introduced their own
versions that offer many similar guarantees, but a deeper look is required to realize just how and why
what Pure Storage has done is different. That is a challenge that the vendor must overcome in the
marketing of its architectural design choices (and the Evergreen Storage program). Existing customers
get it, but if they never move beyond the prospect stage, other enterprises may not see the link
between the vendor's system architecture decisions and the program specifics and how that drives
differentiating value. Like many things in life, the Evergreen Storage program and its architectural tie-
ins present both a challenge and an opportunity for Pure Storage as a vendor.

CONCLUSION

Many enterprise storage vendors have introduced "guarantee" programs that are incrementally changing
the experience of enterprise SAN ownership for the better. But for most of these vendors, they are, in
fact, just programs with no supporting architectural tie-ins in their storage systems. The guarantees that
Pure Storage makes in its Evergreen Storage program are enhanced by specific features of its system
design — namely, software-driven architecture, stateless controllers, modularity of components, and
simplicity of design and operation. Taken together, these capabilities enable simple, nondisruptive
technology refresh at reduced cost throughout the life of a Pure Storage FlashArray and can realistically
extend the life of an array to 10 years without imposing the performance, availability, scalability, or
functional limitations that typically drive the requirement for a forklift upgrade.

Related benefits, like avoiding the need to migrate data when moving to newer storage technologies,
save time and reduce cost while inherently underlining the fairness of only having to pay for storage
capacity once even as a system evolves to higher storage densities through the integration of newer
media technologies. What Pure Storage has done with its approach has raised the bar in customer
expectations around the experience of enterprise storage ownership, and it is one of the reasons why
the company has consistently enjoyed a CX rating (as shown by its high validated Network Promoter
Score) in the top 1% of all businesses across all industries.

The bottom line is that there are specific design features that support Pure Storage's Evergreen
Storage claims and drive a differentiating CX around FlashArray ownership. These features all come
together to define the FlashArray value proposition — that customers will not have to give up
performance to get simplicity of management and growth, efficient usage of storage resources, or a
compelling total cost of ownership for their enterprise storage platforms.

©2020 IDC #US46800220 9

About IDC
International Data Corporation (IDC) is the premier global provider of market intelligence, advisory
services, and events for the information technology, telecommunications and consumer technology
markets. IDC helps IT professionals, business executives, and the investment community make fact-
based decisions on technology purchases and business strategy. More than 1,100 IDC analysts
provide global, regional, and local expertise on technology and industry opportunities and trends in
over 110 countries worldwide. For 50 years, IDC has provided strategic insights to help our clients
achieve their key business objectives. IDC is a subsidiary of IDG, the world's leading technology
media, research, and events company.

Global Headquarters

5 Speen Street
Framingham, MA 01701
USA
508.872.8200
Twitter: @IDC
idc-community.com
www.idc.com

Copyright Notice

External Publication of IDC Information and Data — Any IDC information that is to be used in advertising, press
releases, or promotional materials requires prior written approval from the appropriate IDC Vice President or
Country Manager. A draft of the proposed document should accompany any such request. IDC reserves the right
to deny approval of external usage for any reason.

Copyright 2020 IDC. Reproduction without written permission is completely forbidden.