SanDisk FlashSoft Data Management

Lab Validation

By Vinny Choinski, Senior ESG Lab Analyst, and Kerry Dolan, ESG Research Analyst

Report

SanDisk FlashSoft

Active Data Management

August 2012

Lab Validation: SanDisk FlashSoft 2

Contents

Introduction .................................................................................................................................................. 3

Background ............................................................................................................................................................... 3

ESG Lab Validation ........................................................................................................................................ 5

Getting Started ......................................................................................................................................................... 5

Manageability ........................................................................................................................................................... 7

Application Performance ........................................................................................................................................ 11

Density Scale ........................................................................................................................................................... 13

ESG Lab Validation Highlights ..................................................................................................................... 15

Issues to Consider ....................................................................................................................................... 15

The Bigger Truth ......................................................................................................................................... 16

Appendix ..................................................................................................................................................... 17

ESG Lab Reports

The goal of ESG Lab reports is to educate IT professionals about data center technology products for companies of all types and sizes. ESG Lab reports are not meant to replace the evaluation process that should be conducted before making purchasing decisions, but rather to provide insight into these emerging technologies. Our objective is to go over some of the more valuable feature/functions of products, show how they can be used to solve real customer problems, and identify any areas needing improvement. ESG Lab’s expert third-party perspective is based on its own hands-on testing as well as on interviews with customers who use these products in production environments. This ESG Lab report was sponsored by SanDisk.

All trademark names are property of their respective companies. Information contained in this publication has been obtained by sources The Enterprise Strategy Group (ESG) considers to be reliable but is not warranted by ESG. This publication may contain opinions of ESG, which are subject to change from time to time. This publication is copyrighted by The Enterprise Strategy Group, Inc. Any reproduction or redistribution of this publication, in whole or in part, whether in hard-copy format, electronically, or otherwise to persons not authorized to receive it, without the express consent of The Enterprise Strategy Group, Inc., is in violation of U.S. copyright law and will be subject to an action for civil damages and, if applicable, criminal prosecution. Should you have any questions, please contact ESG Client Relations at 508.482.0188.

Lab Validation: SanDisk FlashSoft 3

Reduced/deferred hardware

Improved total cost of ownership

Improved SLAs

Reduced operational expenses

Improved resource utilization

Reduced power consumption

Improved application performance

Current solid-

Potential

Introduction

This report documents the results of ESG Lab testing of SanDisk FlashSoft for VMware vSphere software. Testing focused on application performance, virtual machine (VM) density scalability, and ease of integration into a VMware environment.

Background

Organizations are increasingly deploying solid-state drives (SSD) in their data centers. Recent ESG research uncovered what benefits SSD users are experiencing as well as what benefits potential users hope to achieve. As Figure 1 shows, improving application performance is at the top of both lists, followed by reducing power consumption, improving resource utilization, and reducing operational expenses.

Figure 1. Benefits of Solid-state Storage, Actual (Current Users) vs. Expected (Potential Adopters)

Which of the following benefits has your organization realized – or does it

expect to realize – as the result of deploying solid-state storage technology?

(Percent of respondents, multiple responses accepted)

1

52%

69%

42%

45%

36%

34%

state storage users (N=91)

36%

35%

32%

23%

solid-state storage adopters (N=132)

30%

(TCO)

31%

22%

capital expenditures

0% 20% 40% 60% 80%

32%

Source: Enterprise Strategy Group, 2011

SSDs can deliver orders-of-magnitude improvements in I/O over traditional hard disk drives (HDDs), offering organizations the opportunity to reduce storage bottlenecks, speed application performance, and minimize latency. While these features are beneficial in physical server environments, they become crucial in virtualized environments in which workloads are aggregated and I/O becomes increasingly randomized. The challenge for many organizations is to know how to apply SSD for maximum benefit. Identifying what applications and data will be “hot” at what time and for how long is difficult at best, and as a result, manually assigning SSD to specific applications or VMs is a hit-or-miss proposition.

1

Source: ESG Research Report, Solid-state Storage Market Trends, November 2011.

Lab Validation: SanDisk FlashSoft 4

SanDisk FlashSoft for VMware vSphere

FlashSoft for VMware vSphere is a software module designed to accelerate application performance in VMware guest machines on a physical server. FlashSoft increases the performance of I/O operations in VMware environments by creating a software-based read cache using virtualized SSD capacity; SSD is dynamically allocated to accelerate virtual disks (VMDKs) as needed. This increase in I/O performance also enables greater virtual machine density. No agents are used in the guest OS, and the software requires less than 10MB of disk space to install and less than 140MB of system memory at runtime. Because FlashSoft is software-based, the cache can consist of any SSD memory device connected via PCIe, SAS, or SATA interfaces, and SSDs can reside in the server or

outside of it. FlashSoft can accelerate data in DAS, SAN, or NAS-based data stores without restriction.

FlashSoft is installed quickly using the standard vSphere Installation Bundle (VIB), and it runs as a block-level file device switch (FDS) loadable module in the ESX kernel (see Figure 2). This software module requires no changes to

the storage configuration, and it operates in the ESX cluster without restrictions. FlashSoft for VMware vSphere is managed through a vCenter console plug-in and supports all native VMware functionality including high availability, vMotion, Storage vMotion, snapshots, clones, linked clones, and VDI. It can be installed in some or all servers in a cluster, and it can accelerate the VMDKs of any servers in which it is installed. Acceleration can selectively be started and stopped on any VMDK individually.

Figure 2. SanDisk FlashSoft Software

FlashSoft accelerates VMDKs on VMFS or NFS storage. When FlashSoft is installed on servers running multiple

virtual machines, the cache space on the SSD is shared dynamically among the VMs. During read activity, all data

passes through the FlashSoft driver in the storage stack; FlashSoft identifies the most frequently used data and

caches it on SSD. In many cases, while only a small percentage of a total data set is “hot,” it can account for 70% to

80% of server I/O. As a result, implementing a relatively small SSD-based cache for frequently accessed data can

provide dramatic performance improvements and eliminate the need to store all of an application’s data on SSD.

Lab Validation: SanDisk FlashSoft 5

ESG Lab Validation

ESG Lab validated prior performance testing and performed remote hands-on evaluation of a FlashSoft

implementation via a lab located at the Milpitas, California, SanDisk facility. Using industry-standard tools and

methodologies, ESG Lab conducted testing that was designed to demonstrate integration with VMware features, ease of management, performance scalability, and the ability to increase virtual machine density.

Getting Started

The test bed consisted of one Dell R810 server with VMware ESX, one Dell R720 server with VMware ESX, and two

Dell R710 servers in an ESX cluster. These hosts were configured with 180GB of LSI WarpDrive PCIe SSD storage. All

servers were connected via iSCSI SAN to a storage array with 24 x 15K RPM SAS drives. The VMware vCenter

Management Interface was connected via LAN to the vSphere server.

Figure 3. FlashSoft Test Bed

ESG Lab Testing

Installing the FlashSoft application into a VMware ESX cluster was extremely simple, and the task was performed

without disruption to the cluster. First, using VMware vMotion, the virtual machines were moved from the first Dell

R710 node to the second. Next, the FlashSoft ESX kernel module was installed on the first host (now containing no

VMs). Following that step, the VMs were moved back to node 1 using vMotion, and FlashSoft was then installed on

the second Dell R710. Service from the cluster was not interrupted at any point during installation. Figure 4 shows

output from the command line interface as install file uncompression and installation was occurring.

Lab Validation: SanDisk FlashSoft 6

Figure 4. Installation Procedure

When the installation was complete, a FlashSoft tab appeared in the vCenter console. Figure 5 shows the vCenter

dashboard view. The “Status of the system” box shows that SSDs and licenses are all set for the ESX cluster. In the

“Utilization” area, cache read hits and misses are beginning to be displayed, and in the SSD area, the device name

and size appear.

Figure 5. VMware vCenter FlashSoft Management

Lab Validation: SanDisk FlashSoft 7

Why This Matters

As virtual servers become a more common implementation for mission-critical applications, they must be able

to support the high levels of application uptime and data availability required. Organizations may avoid

performance-enhancing solutions that are disruptive and require separate management, giving up business

value in order to maintain operations.

ESG Lab validated that FlashSoft for VMware vSphere can be non-disruptively installed in a running VMware

cluster. FlashSoft was easily installed into the ESX kernel in approximately 35 seconds using the standard VIB-

based installer, required no changes to storage configurations, added no footprint on the server, and remained

transparent to existing ESX software layers. Once installed, FlashSoft was available for management from the

vCenter console.

Manageability

The vCenter plug-in enables SanDisk FlashSoft acceleration to be managed directly from VMware. No other

management software is required, and the full functionality of all VMware features is available through a tab in the

vCenter GUI. FlashSoft can also accelerate snapshots at a granular level. The administrator has the option to

accelerate a base (primary) data set, individual point-in-time snapshots, or both; this level of flexibility enables the

administrator to focus acceleration wherever needed in order to optimize the performance impact.

ESG Lab Testing

ESG Lab tested the impact of FlashSoft installation on vMotion capabilities. The top picture in Figure 6

demonstrates vMotion in the cluster before FlashSoft was installed; ESG Lab moved a running VM from the first ESX

node in the cluster to the second node. The bottom picture shows that vMotion continued to operate normally

after FlashSoft was installed.

Lab Validation: SanDisk FlashSoft 8

Figure 6. VMware vMotion with FlashSoft

FlashSoft can accelerate snapshots to speed data management activities. While other solutions may accelerate

performance only for the primary dataset or the entire volume, FlashSoft can apply acceleration to individual

snapshots as well as the base dataset. Figure 7 shows the granularity with which FlashSoft can apply acceleration;

any primary copy or snapshot in the tree structure can be designated for acceleration.

Lab Validation: SanDisk FlashSoft 9

Figure 7. Acceleration Target Selection

The FlashSoft tab also enables administrators to monitor performance from within vCenter. Read hits, misses, and

operations are displayed graphically, with some statistics listed in table format as well (see Figure 8).

Figure 8. Performance Monitoring

Lab Validation: SanDisk FlashSoft 10

Why This Matters

As virtualization expands in a data center, the traditionally separate infrastructure domains begin to converge.

Suddenly, instead of a clear definition of tasks assigned to each domain, the server administrator may be taking

actions that affect storage, and separate management tools can add complexity and mask a holistic view. When

ESG asked virtualization users what developments would enable more widespread use of server virtualization,

application, server, storage, network, and security personnel mentioned better integration of tools across

infrastructure domains and the need for virtualization-aware management tools.2

ESG Lab validated the ease of managing FlashSoft acceleration from within the vCenter console and confirmed that

vMotion operated normally after FlashSoft implementation. Administrators can view the FlashSoft performance

chart from within vCenter, making acceleration details easily accessible alongside other virtual server management

information.

ESG Lab also confirmed that FlashSoft can accelerate primary data sets, individual snapshots, or both to improve

performance. This feature is particularly beneficial for improving business productivity for test and development,

backup, VDI, and other multi-copy use cases. Accelerating test snapshots can enable more test iterations to be

completed in less time, clone acceleration can keep virus scans from slowing performance, and accelerating

snapshots for backup can prevent data protection from interrupting business operations.

2

Source: ESG Research Report, The Evolution of Server Virtualization, November 2010.

Lab Validation: SanDisk FlashSoft 11

Application Performance

Application performance was measured by comparing the performance of VMs running an OLTP test workload on

an unaccelerated baseline system to an accelerated system using FlashSoft software with SSD. The OLTP workload

was tested on a pair of VMs using HDDs only; then it was compared to the same pair of VMs with FlashSoft

software and an SSD read cache target added. Typical database performance metrics were captured and analyzed

for both configurations.

ESG Lab Testing

ESG Lab began performance testing by running an OLTP workload on two VMs configured with HDD data stores.

The data stores hosted the OLTP application. Testing was run in an HDD-only configuration and then repeated with

FlashSoft-managed SSD acceleration enabled. Figure 9 shows the transactions-per-minute performance benefit of

FlashSoft acceleration over the HDD-only configuration.

Figure 9. OLTP Transactions per Minute

What the Numbers Mean

• The OTLTP application was configured with a 4,500 customer database.

• The OLTP application was configured with 95 connections.

• The OLTP application with FlashSoft was able to achieve an average of 4x more transactions per minute

than the HDD-only configuration.

Lab Validation: SanDisk FlashSoft 12

Next, ESG Lab explored the effects of acceleration on database batch requests. Using the same OLTP workload

described in the transactions-per-minute testing, ESG Lab captured and analyzed average batch request numbers

before and after acceleration. Figure 10 shows the average batch requests per second for the same pair of VMs

with an HDD-only configuration compared with adding FlashSoft acceleration with SSD.

Figure 10. Database Batch Requests

What the Numbers Mean

• The FlashSoft configuration was able to handle 3.5x more batch requests per second.

Why This Matters

By aggregating multiple workloads on physical servers, virtualization often increases the randomization of I/O,

which can create storage bottlenecks that affect application performance. While performance problems are never

welcome, the increasing consumerization of IT raises the expectations of users and decreases their tolerance for

slowing of application and data access. In addition, many organizations resist placing tier-1 applications on virtual

servers for fear that workload aggregation will slow performance. ESG research on virtualization revealed that after

budget concerns and lack of legacy application support, performance issues were the key concern preventing

organizations from expanding their virtualization deployments.3

ESG Lab validated the ability of FlashSoft for VMware vSphere to accelerate the transaction performance of a

MySQL database by 4x over an HDD-only solution. ESG believes that this level of performance improvement can

enable organizations to migrate more business-critical applications to virtual machines without hindering

application performance or data access.

3

Source: ESG Research Report, The Evolution of Server Virtualization, November 2010.

Lab Validation: SanDisk FlashSoft 13

Density Scale

Density scale is the ability to increase the number of virtual machines running on a physical server while still

delivering the performance required to efficiently run a mix of workloads typically encountered in current

enterprise compute environments. For density testing, ESG Lab leveraged a number of workloads, including an

application server, web server, mail server, and supporting database and infrastructure servers. The servers were

grouped together in tiles containing six servers each. The performance was measured on the application, web and

mail servers, as the number of tiles was scaled.

ESG Lab Testing

ESG Lab began density testing by simply measuring the number of VMs that could be run on a given server before

and after acceleration, while maintaining manageable response times for each application. Figure 11 shows the

total number of VMs (24 vs.72) for each scenario.

Figure 11. VM Density

What the Numbers Mean

• The mixed-workload environment was able to support 48 more VMs with FlashSoft vs. HDD only.

• Only 24 VMs could be supported in the HDD configuration before poor response times were observed.

• FlashSoft Improved VM density scale by 3x.

Next, ESG Lab measured the ability of the environment to handle an increased number of requests by each

application with FlashSoft acceleration enabled. Figure 12 shows the total number of requests as well as the

number of requests per application, before and after acceleration.

Lab Validation: SanDisk FlashSoft 14

Figure 12. VM Workload

What the Numbers Mean

• FlashSoft achieved a performance gain of 3x more requests per second than the HDD-only configuration.

• All applications demonstrated a significant and fairly evenly balanced performance gain.

Why This Matters

Storage I/O performance in a virtual environment affects more than just application speed. Faster IOPS enable

organizations to increase the virtual-to-physical server consolidation ratio—that is, the VM density on each physical

machine. A lack of adequate performance inhibits the consolidation ratio, which directly affects the potential

savings available from consolidation as well as constraining virtualization expansion. If you can scale the number of

applications on each host while maintaining performance SLAs, you can scale the application capacity of your data

center without adding hardware. This approach not only enables greater data center productivity without

increasing capital and operational expense, but it also increases the opportunity to virtualize tier-1 applications.

Using a tile-based benchmark methodology, ESG Lab validated the VM density scalability that FlashSoft for VMware

vSphere enabled. The testing demonstrated scalability to 72 VMs (a 3x improvement), which was able to satisfy

more than 2,100 requests per second for a combination of application, web, and e-mail applications.

Lab Validation: SanDisk FlashSoft 15

ESG Lab Validation Highlights

 SanDisk FlashSoft was installed easily and non-disruptively in approximately 35 seconds into an ESX host,

with status and SSD utilization details immediately available in a new vCenter tab.

 ESG was able to accelerate both base data and snapshot targets by selecting objects from a device list for

each VM. We were pleased to see the ability to accelerate not only the primary data copy, but also snapshot copies. This can be extremely useful in test and development and VDI environments.

 ESG Lab was able to scale the VM density of a mixed-workload environment to 72 virtual machines on a

single Dell R810 server with FlashSoft. With an HDD-only configuration, the environment could only scale to 24 virtual machines. This represents a 3x increase in VM density.

 ESG confirmed the ability of FlashSoft to improve OLTP application performance significantly. We observed

a 4x improvement in transactions per minute and a 3.5x improvement in batch request processing over the same configuration with HDD-only devices.

Issues to Consider

 Although management was well integrated into VMware vCenter and no agent had to be installed on the

guest machine to utilize FlashSoft software, VMs marked for acceleration needed to be restarted to initiate caching. ESG Lab hopes to see the requirement removed in a future release of the product.

 To improve support for VDI environments, ESG Lab believes the ability to implement write caching with

FlashSoft would greatly enhance its ability to address other challenging virtual infrastructure I/O situations. The lab hopes to see the feature incorporated in a future release.

Lab Validation: SanDisk FlashSoft 16

The Bigger Truth

Both server virtualization and solid-state disk deployments appear to be moving into the next wave of greater

customer adoption. Server virtualization has clearly demonstrated significant business value in reducing both OPEX

and CAPEX, and as a result has reigned for three years running as the top priority in ESG’s annual IT spending

intentions survey.

applications; in addition, they are leveraging virtualization to gain mobility and agility across the enterprise.

However, the workload aggregation of virtual server environments can cause storage I/O bottlenecks in traditional

storage environments by creating an “I/O blender” effect—multiple types and sizes of workloads can jam disk

spindles by overloading the number of IOPS they can handle. This overload not only affects application

performance, but also restricts the ability to increase VM density for both virtual deployment expansion and

greater savings. The performance problem is a show-stopper that prevents many organizations from moving

business-critical applications to virtual machines.

Solid-state disk usage is also on the rise—ESG research with North American IT professionals conducted in 2011 indicated that more than one-third (34%) of respondents were already using SSD, while 17% had plans to add SSD within the year, and another 18% were in the evaluation phase. Only 17% of respondents had no interest or familiarity with the technology.

adoption, with performance the primary benefit.

SSD can improve I/O performance in virtualized environments, but manual placement of SSD by application is

unreliable. At the same time, dedicating entire arrays to SSD can be not only expensive, but also wasteful; a full SSD

array is overkill when only 10% to 20% of data stores need that level of performance. In addition, many

organizations cannot add SSD capacity to their storage implementations because of incompatibility, and they

cannot deploy an SSD-compatible storage platform without jeopardizing current efforts related to compliance, data

security, backup, disaster recovery, etc.

4

Customers have moved past the initial efficiency gains and are virtualizing more mission-critical

5

This research indicated that cost remains the primary impediment to solid-state

SanDisk offers a way out of that morass with its FlashSoft for VMware vSphere software. FlashSoft provides

organizations with the opportunity to take advantage of SSD performance on the server side, managed

automatically, while they maintain their existing storage implementations. ESG Lab found FlashSoft for VMware

vSphere easy to install and manage through the vCenter console. Using FlashSoft software-based read cache with

heterogeneous SSDs, ESG Lab validated not only a significant increase in application performance, but also a

significant increase in virtual machine scalability. FlashSoft was easy to install and was able to accelerate workloads without restricting any VMware features.

By improving I/O performance without manual intervention, SanDisk FlashSoft can enable organizations to expand their virtual server deployments while speeding application performance, and still gain the benefits of consolidation that translate into both capital and operational cost savings. The ease of implementing FlashSoft in a VMware environment makes the proposition especially appealing.

4

Source: ESG Research Report, 2012 IT Spending Intentions Survey, January 2012.

5

Source: ESG Research Report, Solid-state Storage Market Trends, November 2011.

Lab Validation: SanDisk FlashSoft 17

Model: R720

Model: R810

Model: R710 Model: MD3220 Model: MD3220

PCI SSD Controller

LSI WarpDrive PCIe

VMware

Version: ESXi 5.0.0

Web Server

Apache 2.2.15

Application Server

Oracle GlassFish Server v3.1.1

Mail Server

Dovecot 2.0.9

Database Server

PostgresSQL 9.1.3

Appendix

Table 1. ESG Lab Test Bed

Dell PowerEdge R720

Dell PowerEdge R810

Dell PowerEdge R710

Servers

CPU: Intel Xeon E5-2660 Processor Cores: 16 cores Processor Speed: 2.2 GHz RAM: 128GB

CPU: Intel Xeon E7-4860 Processor Cores: 40 cores Processor Speed: 2.27 GHz RAM: 256GB

CPU: Intel Xeon 5500 Processor Cores: 4 Processor Speed: 2.2 GHz RAM: 48GB

Dell PowerVault

Storage

Drives: 24 x 300GB 10K RPM Type: 6Gb/sec SAS RAID Level: RAID-0

Drives: 24 x 146GB 15K RPM Type: 6Gb/sec SAS

Applications and Operating Systems

20 Asylum Street | Milford, MA 01757 | Tel: 508.482.0188 Fax: 508.482.0218 | www.esg-global.com