Buyer’s Guide: How to Evaluate All-Flash Storage · - Computational research division, U.S. national lab 4.1 Sustained Low Latency Although many all-flash storage systems can support

TABLE OF CONTENTS

Abstract The technical and financial advantages of solid-state drives are driving a rapid transition away from traditional disk and hybrid flash storage to all-flash arrays (AFAs) for most primary workloads. This white paper identifies some of the key considerations for evaluating all-flash storage systems based on your application environment, including virtualized applications, next-generation cloud services, and dedicated mission-critical applications.

White Paper

Buyer’s Guide: How to Evaluate All-Flash Storage July 2016 | WP-7234

2 Buyers Guide: How to Evaluate All-Flash Storage © 2016 NetApp, Inc. All Rights Reserved.

1 The All-Flash Tipping Point ............................................................................................................. 3

1.1 Evaluating All-Flash Storage ....................................................................................................................... 3

2 Virtualized Applications .................................................................................................................. 4

2.1 Enterprise Data Management ..................................................................................................................... 4

2.2 Nondisruptive Operations ........................................................................................................................... 4

2.3 Application Integration ................................................................................................................................ 4

3 Next-Generation Cloud Services .................................................................................................... 5

3.1 Granular Scale-Out ..................................................................................................................................... 5

3.2 Guaranteed Performance ........................................................................................................................... 5

3.3 Self-Healing Design .................................................................................................................................... 5

4 Latency-Sensitive Applications ...................................................................................................... 6

4.1 Sustained Low Latency ............................................................................................................................... 6

5 Additional Considerations .............................................................................................................. 6

5.1 Vendor Performance Claims ....................................................................................................................... 6

5.2 Proof-of-Concept Testing ............................................................................................................................ 6

5.3 Efficiency Guarantees ................................................................................................................................. 7

5.4 Total Cost of Ownership ............................................................................................................................. 7

5.5 Extended Warranties .................................................................................................................................. 7

6 NetApp All-Flash Solutions ............................................................................................................. 8

6.1 NetApp AFF ................................................................................................................................................ 8

6.2 NetApp SolidFire ......................................................................................................................................... 8

6.3 NetApp EF-Series ....................................................................................................................................... 9

For More Information ............................................................................................................................. 9

LIST OF FIGURES Figure 1) Criteria for evaluating all-flash arrays .............................................................................................................. 3Figure 2) NetApp all-flash solutions ................................................................................................................................ 8


1 The All-Flash Tipping Point Flash storage was originally introduced as a premium performance tier. By substituting solid-state performance for mechanical disk drives, you could quickly and easily accelerate performance for dedicated databases and virtual desktop environments. Flash was also widely used to increase the input/output operations per second (IOPS) for hybrid systems that used both solid-state drives (SSDs) and spinning disk drives. However, the adoption of all-flash systems remained limited because SSD capacity was simply more expensive than disk capacity.

More recently, though, the price of flash media has dropped dramatically. At the same time, process improvements in flash component design have enabled more information to be crammed into each device, leading to improved SSD density. The physical capacities of SSDs have now surpassed those of the largest hard disk drives, and the capacity advantage is expected to keep increasing for the foreseeable future. When combined with inline storage efficiency techniques, such as deduplication and compression, the cost of effective capacity (the amount of logical capacity after storage efficiencies are applied) for all-flash systems is now on par with that of disk systems, and all-flash systems have proven far superior to disks when comparing operational costs for power, cooling, space requirements, and management.

Malcolm Gladwell described the rapid adoption of powerful new ideas, products, and norms when conditions are right and labeled it the “tipping point” in his best-selling book of the same name. A tipping point requires an environment ready for change. Today’s enterprise IT environments, which are characterized by a scarcity of resources and economic pressure to control costs, are ready for change at precisely the same time that flash technology is making rapid improvements. As a result, all-flash arrays have reached a tipping point and are now replacing both disk and hybrid flash architectures for an increasing range of primary IT workloads.

1.1 Evaluating All-Flash Storage When evaluating all-flash storage systems, it helps to start by considering your application environment. These tend to fall into one of three categories, each of which has requirements that call for a specific type of all-flash solution:

• Virtualized applications that share a common infrastructure • Next-generation services designed for a cloud infrastructure • Dedicated applications that require extremely low latency

Figure 1) Criteria for evaluating all-flash arrays.


2 Virtualized Applications Many all-flash arrays have been designed to accelerate dedicated applications, such as Microsoft or Oracle databases. These designs can work well for the target applications, but they lack some of the basic features you need to consolidate virtualized applications onto a shared storage infrastructure, such as an ability to support NAS-based file systems or replicate data to alternate storage tiers. This creates a number of inefficiencies, because:

• The dedicated arrays must be managed separately. • Performance and capacity are siloed and cannot be shared across workloads. • Protection copies cannot be stored on lower-performing, lower-cost storage tiers.

The following sections highlight some of the basic features required to optimize the value of all-flash solutions when they are used to consolidate virtualized applications onto a single storage platform.

“We can support the business with much more agility now that we have virtualized our most critical Oracle and Microsoft SQL Server databases on all-flash storage.”

- Lead for storage and compute platforms, financial services company

2.1 Enterprise Data Management In general, enterprise data management encompasses the features and data services that previously enabled disk and hybrid flash arrays to serve as a consolidation point for a wide range of virtualized applications. In order to consolidate these applications on all-flash storage, the same capabilities are needed, including:

• Support for NFS and SMB/CIFS file access in addition to FC and iSCSI block services • Integrated data protection, including local and remote point-in-time copies • Consistent high performance with storage efficiency features enabled and active • Quality of service (QoS) controls and secure partitioning for multitenant environments • Support for business continuity over wide area and metropolitan area networks • Single-interface management, along with API and command-line options for custom automation

As IT organizations increase their use of cloud services to increase agility and reduce costs, there is also a growing need to move workloads and data from on-premises systems to external services and vice versa. All-flash systems that support a hybrid cloud design provide you with the flexibility to incorporate cloud services for workloads such as dev/test, backup and recovery, and disaster recovery as needed.

2.2 Nondisruptive Operations Nondisruptive operations provide the ability to perform maintenance activities and reconfiguration operations, including the addition or removal of controller nodes and disk shelves, without interrupting storage services to hosts and clients. This capability also enables the rebalancing of performance and capacity across a storage platform without the need to schedule application downtime.

2.3 Application Integration Out-of-the-box integration with enterprise applications, such as those from Citrix, Microsoft, Oracle, VMware, and SAP, enables predictable and reliable provisioning and protection of application-consistent copies of your data. Without this capability, IT teams are forced to invest their limited time and resources developing and supporting custom scripts for each application.


3 Next-Generation Cloud Services Next-generation cloud services are designed to run in the public cloud or on top of a cloudlike infrastructure where software is delivered as a service, either to your internal users or to external customers.

These environments require scale, agility, and economics similar to those of public cloud infrastructures run by Amazon or Google. Although these environments require many of the core all-flash capabilities as virtualized application environments, they also require additional features, including granular scale-out, guaranteed performance, and self-healing design.

“We figured there’s got to be a better way to do storage.” - Director of global platform engineering, cloud services provider

3.1 Granular Scale-Out Scale-out has become the predominant storage architecture for the deployment of cloud services because of its ability to offer seamless and transparent resource expansion without the cost and complexity of traditional infrastructure. In a scale-out design, pools of storage are shared resources. The ability to cluster these resources provides incremental, on-demand scaling. IT teams can purchase and manage additional storage resources, such as capacity and IOPS, as granularly as one node at a time based on business needs.

A granular scale-out architecture also provides the flexibility to independently and nondisruptively scale both capacity and performance in a predictable linear pattern. This means that businesses can distribute data and traffic across all of the nodes in a cluster as the scope of their data services changes over time.

3.2 Guaranteed Performance In a next-generation cloud infrastructure, QoS controls are enforced by allocating a guaranteed amount of storage performance to each application. Every application provisioned is assigned a guaranteed range of IOPS, and those IOPS levels are then maintained, regardless of any other application activity, capacity level, or I/O pattern. Each application receives consistent, predictable performance based on the requirements of and value to the business, while isolating and protecting the workload from every other workload and from hardware/software faults.

The ability to guarantee firm performance SLAs for applications, workloads, and tenants across a storage infrastructure can be a key differentiator for cloud service providers and enterprise clouds alike.

3.3 Self-Healing Design Deploying an all-flash storage architecture at cloud scale typically introduces a series of challenges, such as how to automatically recover from failed controllers or media. Recovering from a failure with a traditional RAID-based storage design can take hours or days, which can affect application performance and data availability. Because of this, cloud services that need to scale applications with guaranteed levels of performance often prefer replication-based high availability (HA) instead of traditional RAID data protection.

Replication-based HA is a post-RAID data protection scheme based on a distributed replication algorithm. This architecture protects against both drive and node failure. It allows the storage system to absorb concurrent failures across all levels of the storage solution. In a failure event, each drive in the system redistributes a small percentage of its data in parallel to the free space on all other remaining drives, and no operator intervention is required. The self-healing properties of an automated, replication-based architecture enable fast rebuilds with minimal performance impact. With this architecture, faults are isolated, and QoS settings remain enforced.


4 Latency-Sensitive Applications For some applications, such as mission-critical online transaction processing (OLTP) and online analytics processing (OLAP), what matters the most is extreme performance. Anything that slows down the I/O path to the storage media is considered unacceptable. Even the small amount of overhead introduced by inline storage efficiencies can introduce delays and cause problems for these applications.

These applications typically require dedicated hardware in order to deliver the highest levels of performance and often run on bare metal servers rather than virtual servers due to their requirements for speed. In many cases, these applications have their own methods for addressing data protection, storage efficiency, and disaster recovery and do not need or depend on array-based data management services.

“The biggest challenge for us is the analysis of the data in as near real time as possible.” - Computational research division, U.S. national lab

4.1 Sustained Low Latency Although many all-flash storage systems can support latencies as low as 1ms, applications that require extreme performance require consistent, sustained response times that are measured in hundreds of microseconds rather than milliseconds. These applications do not tolerate latencies that fluctuate due to backend storage services or that increase rapidly as the IOPS load increases. The reason is simple: revenue, customer satisfaction, or service-level penalties are typically at risk for applications that fall into this category.

5 Additional Considerations Flash technology has introduced a new set of options for IT organizations. By deploying all-flash systems, you can take advantage of increased and faster throughput to significantly boost performance for the majority of your business applications. However, in addition to selecting an all-flash solution that provides the best fit for your application environment, there are several other considerations that should be factored into your all-flash section process.

5.1 Vendor Performance Claims Flash is ideal for optimizing performance, but there can be wide performance variations across vendor systems, and even wide variability within a single system, depending on a number of factors. In general, you should seek solutions that can demonstrate consistent, low latency under 1ms based on third-party benchmarks that simulate real-world workloads, such as SPC-1 and TPC-E. You should also take claims about “maximum IOPS” with a grain of salt unless a vendor can provide you with more details. For example, without the answers to the following questions, you are unable to make a valid apples-to-apples comparison across competing solutions:

• What was the I/O size used to run the test? • How does this compare with the I/O size for my applications? • What was the read/write mix? • Were the I/Os random or sequential? • Most important of all: What was the latency at increasing IOPS intervals, and at what IOPS level did

the latency exceed 1ms?

5.2 Proof-of-Concept Testing Although benchmark data is useful for making comparisons, the real litmus test is to run an all-flash solution using your own applications and data. Vendors typically offer a limited time test drive program, and some even let you use their own facilities to run sophisticated proof-of-concept (POC) tests.


However, beware of vendors that use sales tactics to try and avoid a hands-on evaluation or that insist on using nonstandard configuration settings or their own special datasets to demonstrate performance or storage efficiency savings. These are all red flags that a vendor’s performance claims may not be achievable in a real-world environment.

5.3 Efficiency Guarantees When evaluating all-flash systems, you may come across storage efficiency claims that are simply too good to be true. If you’re unable to verify (or disprove) such claims by testing with your own applications and data, you should request that your vendor back up the claims in writing, including the remediation that is provided to your company for failure to deliver as promised. Reputable vendors back up their guarantees in writing.

5.4 Total Cost of Ownership In order to compare the total cost of ownership (TCO) of an all-flash array versus disk-based or hybrid flash systems, the following items should be considered:

• Effective capacity. This is the amount of logical capacity available to store application data after storage efficiencies are applied. At a minimum, your calculations should include the inline efficiencies available with an all-flash system, such as data compression and deduplication. You should also include Snapshot® copies and clones if they are recommended as best practices for your particular use case.

• Application integration. The management costs for provisioning, protecting, and recovering application data are often overlooked when evaluating primary storage. However, the capabilities and costs for managing application-consistent data copies can vary widely across all-flash offerings and should be included when calculating your TCO.

• Data center operational costs. Power, cooling, and data center real estate are all important operational costs that must be considered. According to one user survey, all-flash storage systems can provide over 70% savings in power and cooling alone when compared with disk-based systems.

• IT management. When storage performance no longer needs to be treated as a scarce resource, IT personnel are freed up across the entire organization. Database, application, and infrastructure specialists can all reclaim time that was previously spent architecting and managing storage performance.

• Software licenses. The consolidation of workloads onto an all-flash platform can also generate software savings. Database software licenses and maintenance fees can be reduced when all-flash storage systems can deliver the desired level of performance while using fewer CPU cores or servers.

5.5 Extended Warranties Some IT professionals continue to be concerned about the durability of flash media, or flash “wear-out,” even though there is now a wealth of real-world data demonstrating that SSDs and the all-flash systems that incorporate them are extremely reliable. The good news is that most vendors are so confident in the long-term reliability of their all-flash systems that they’re happy to provide you with an extended warranty. Other options, such as free controller upgrades, may also be available if you are willing to commit to a long-term service contract.

Remember that you’ll need to factor all of the costs into your total cost of ownership calculations in order to make an apples-to-apples comparison across vendors. For example, you may discover that the potential savings from some of the “free” controller upgrade options are more than offset by higher annual service charges.


6 NetApp All-Flash Solutions The NetApp portfolio of all-flash solutions consists of three product families, each of which is optimized to deliver maximum benefits for a different type of application environment: general-purpose virtual infrastructure, next-generation cloud infrastructure, and dedicated infrastructure that needs to deliver extreme performance.

Figure 2) NetApp all-flash solutions.

6.1 NetApp AFF NetApp All Flash FAS (AFF) systems provide a high-performance solution for consolidating multiple SAN and NAS workloads on a single, unified architecture. With AFF, your investment is protected if your performance and capacity needs change or if your IT strategy evolves to encompass data management across a hybrid cloud.

NetApp AFF systems enable you to:

• Consolidate a wide range of workloads on a single platform, with support for millions of IOPS and hundreds of petabytes of effective capacity across a single cluster

• Eliminate performance silos and seamlessly integrate with hybrid flash and disk systems in an ONTAP® environment, enabling multiple performance tiers

• Transparently move workloads to the storage tier that best meets your applications’ requirements

6.2 NetApp SolidFire SolidFire® systems combine the performance and economics of all-flash storage with a web-scale architecture that radically simplifies data center operations and enables rapid deployments of new applications. These systems are optimized for virtualized cloud and service provider environments where high performance, workload-level QoS control, and agile scalability are top-level requirements. SolidFire systems provide:

• The ability to quickly scale in increments as small as 1U • Control over minimum, maximum, and burst QoS settings so you can guarantee performance to

individual workloads • A software-defined architecture that can be fully automated and supports commodity, off-the-shelf

hardware


6.3 NetApp EF-Series The EF-Series all-flash arrays are designed specifically for high-speed transactional applications that demand high IOPS and consistent low latency. A single EF-Series all-flash system in a 2U enclosure can deliver the performance of over one thousand 15K RPM disk drives while requiring just a fraction of the rack space, power, and cooling. The extreme performance provided by EF-Series arrays can:

• Deliver faster, actionable results from OLTP and OLAP environments • Significantly improve the performance of data analytics applications • Improve the user experience for customer-facing and decision support systems

For More Information • SolidFire Resource Center

http://www.solidfire.com/resources • NetApp All Flash FAS Overview (TR-4505)

http://www.netapp.com/us/media/tr-4505.pdf • Introduction to NetApp EF560 Flash Array (TR-4371)

http://www.netapp.com/us/media/tr-4371.pdf • NetApp Customer Case Studies

http://www.netapp.com/us/company/customer-stories/ • Storage Performance Council SPC-1 Benchmark Results for EF-Series and AFF Systems

http://www.storageperformance.org/results/benchmark_results_spc1_active/#a00173


Refer to the Interoperability Matrix Tool (IMT) on the NetApp Support site to validate that the exact product and feature versions described in this document are supported for your specific environment. The NetApp IMT defines the product components and versions that can be used to construct configurations that are supported by NetApp. Specific results depend on each customer’s installation in accordance with published specifications.

Trademark Information NetApp, the NetApp logo, Go Further, Faster, AltaVault, ASUP, AutoSupport, Campaign Express, Cloud ONTAP, Clustered Data ONTAP, Customer Fitness, Data ONTAP, DataMotion, Flash Accel, Flash Cache, Flash Pool, FlashRay, FlexArray, FlexCache, FlexClone, FlexPod, FlexScale, FlexShare, FlexVol, FPolicy, GetSuccessful, LockVault, Manage ONTAP, Mars, MetroCluster, MultiStore, NetApp Fitness, NetApp Insight, OnCommand, ONTAP, ONTAPI, RAID DP, RAID-TEC, SANshare, SANtricity, SecureShare, Simplicity, Simulate ONTAP, SnapCenter, SnapCopy, Snap Creator, SnapDrive, SnapIntegrator, SnapLock, SnapManager, SnapMirror, SnapMover, SnapProtect, SnapRestore, Snapshot, SnapValidator, SnapVault, SolidFire, StorageGRID, Tech OnTap, Unbound Cloud, WAFL, and other names are trademarks or registered trademarks of NetApp Inc., in the United States and/or other countries. All other brands or products are trademarks or registered trademarks of their respective holders and should be treated as such. A current list of NetApp trademarks is available on the web at http://www.netapp.com/us/legal/netapptmlist.aspx. WP-7234-0616

Copyright Information Copyright © 1994–2016 NetApp, Inc. All rights reserved. Printed in the U.S. No part of this document covered by copyright may be reproduced in any form or by any means—graphic, electronic, or mechanical, including photocopying, recording, taping, or storage in an electronic retrieval system—without prior written permission of the copyright owner.

Software derived from copyrighted NetApp material is subject to the following license and disclaimer:

THIS SOFTWARE IS PROVIDED BY NETAPP “AS IS” AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, WHICH ARE HEREBY DISCLAIMED. IN NO EVENT SHALL NETAPP BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

NetApp reserves the right to change any products described herein at any time, and without notice. NetApp assumes no responsibility or liability arising from the use of products described herein, except as expressly agreed to in writing by NetApp. The use or purchase of this product does not convey a license under any patent rights, trademark rights, or any other intellectual property rights of NetApp.

The product described in this manual may be protected by one or more U.S. patents, foreign patents, or pending applications.

RESTRICTED RIGHTS LEGEND: Use, duplication, or disclosure by the government is subject to restrictions as set forth in subparagraph (c)(1)(ii) of the Rights in Technical Data and Computer Software clause at DFARS 252.277-7103 (October 1988) and FAR 52-227-19 (June 1987).

Documents

Buyer’s Guide: How to Evaluate All-Flash Storage · - Computational research division, U.S. national lab 4.1 Sustained Low Latency Although many all-flash storage systems can support