Performance Characterization of Iomega StorCenter ix4-200r in a Virtualized Database Environment
Introduction Reference Architecture Test Methodology Performance Conclusion

Introduction

Small- and medium-sized businesses (SMBs) typically face the challenge of cost-effectively managing their IT infrastructure that supports a few business-critical applications. These businesses look for hardware and software solutions that offer a combination of low Total Cost of Ownership (TCO), reliability, and performance. Virtualization is a proven software technology that is rapidly transforming the IT landscape and fundamentally changing the way that people compute. VMware virtualization technology simplifies IT so that companies can more effectively utilize their storage, network, and computing resources to control costs and respond faster. VMware ESX Server goes beyond simple server consolidation; it is a proven platform for running mission-critical applications, such as a database at near-native performance. The Iomega StorCenter Pro ix4-200r NAS Server is a unified NAS and IP SAN storage device designed to serve SMBs. The affordable storage device, built on EMCs world-class enterprise storage functionality, offers great reliability, ease of use, and ease of management characteristics. Powered by the EMC LifeLine software, the Iomega ix4-200r delivers economical file sharing, fast data throughput, cross-platform support and enterprise level features to SMBs and remote work groups. Iomega ix4-200r is a certified NAS and iSCSI storage platform for VMware ESX Server. Dell helps businesses transition to virtualization by engineering architected, pre-built configurations that are designed to build a dynamic and efficient virtualized environment from the server and hypervisor, to network storage systems and management toolsets. Dells Business Ready Configurations simplify the design, ordering and deployment of virtual infrastructure and ensure integration completeness and production readiness. This white paper is based on testing conducted jointly by EMC and Dell. The Microsoft SQL Server database was deployed on a VMware ESX Server, a Dell PowerEdge 1950, and an Iomega StorCenter ix4-200r to simulate a simple virtualized environment. An open source utility was used to generate database workloads. Test results prove that the Iomega ix4-200r is an ideal candidate for providing network shared storage to VMware virtual infrastructure while satisfying production application performance requirements in an SMB or ROBO deployment.

Reference architecture

Overview
For the VMware virtualized environment, a Dell PowerEdge 1950 server was used to create a VMware HA cluster. An Iomega StorCenter ix4-200r storage device was used to provide iSCSI shared storage to the VMware ESX Server. The Iomega storage device had four 7200RPM SATA drives built in a RAID 5 configuration. The Dell servers and the Iomega storage device were all connected to a Dell PowerConnect 5448 Ethernet switch. Each Dell server had four NIC connections to the switch. Two of the NICs were teamed together to provide network HA for VMware ESX storage connectivity. The Iomega storage device had a single Gigabit link to the switch. The virtual machine that served as the database server was created using VMFS datastore over iSCSI. A second virtual disk, also using VMFS datastore over iSCSI, was added to the virtual machine to hold SQL Server database files separately from the VM system disk. Another virtual machine that ran the DVD Store load driver application was installed on the ESX host local disk to generate database I/Os.

Page 2 of 8

Iomega Technical Brief

Architecture diagram

Figure 1 Architecture diagram

Hardware resources

Table 1 shows the equipment that was used in the testing.
Table 1. Hardware resources
Description IomegaStorCenterix4200r
Configuration Four7.2KRPM500GBSATAdriveswithRAID5 configuration,sharedstoragepresentedover iSCSI. VMwareESX3.5Update4Host XeonE54603.16GHzQuadCoreProcessor 8GB 12MBL2 1333MHz 2x73GB15KSAS(RAID1) 4xGigabitEthernetadapters
2DellPowerEdge1950 CPU Memory Cache MemoryBus Storage NIC

3 of 8

Software resources
Table 2 shows the software and application versions used on the storage and servers.
Table 2. Software resources

Description EMCLifeline

Configuration Release2.0.15.43099
VMwareESXServer VirtualCenterManagementServer Virtualmachine#1 OS VirtualCPU Memory VirtualDisks Databaseserver Virtualmachine#2 OS VirtualCPU Memory VirtualDisk Databaseloadgenerator
3.5Update4 2.5Update4 Databaseserver WindowsServer2003EnterpriseSP2x1GB4GB 2(C:15GBsystemdiskE:6GBdatabasefiles) MicrosoftSQLServer2005SP3 Loadgenerator WindowsServer2003EnterpriseSP2x1GB 1(C:15GBsystemdisk) DVDStore2.0

Storage layout

Figure 2 shows the storage configuration on the Iomega ix4-200r device and storage distribution on the ESX server hosts and virtual machines. The four SATA drives on the Iomega device were configured into a 3+1 RAID 5 storage group. A single file system was created across the RAID group to utilize all four disk spindles. An iSCSI LUN was created inside the file system and connected to an ESX host using the iSCSI protocol. The VMware ESX hosts connected to the LUN to create a VMFS datastore. The datastore was shared storage between the ESX hosts. The virtual machine serving as the database server was created in such a way that the guest OS and SQL Server database were installed on two separate virtual disks that both resided in the same iSCSI VMFS datastore. The virtual machine serving as the load generator was created differently; it had only one virtual disk that resided in the local VMFS datastore.

4 of 8

Figure 2 Storage layout

Test methodology

Objective
The primary objective of the testing was to observe how the Iomega ix4-200r storage device performs under a random database workload. The test results would help determine whether the device is capable of serving storage via the iSCSI protocol to a small virtualized database deployment.

Test tools

The open source DVD Store application was used to generate a database workload. DVD Store Version 2 (DS2) is a simulation of an online e-commerce application, consisting of a backend database component, a web application layer, and client load-driver components. The goal in designing the database component as well as the mid-tier application was to utilize many advanced database features (transactions, stored procedures, triggers, referential integrity) while keeping the database easy to install and understand. The DS2 workload may be used to test databases or as a stress tool for any purpose. DS2 comes in 3 standard sizes. The Medium database profile was used in this test.

5 of 8

Database Small Medium Large Size 10MB 1GB 100GB Customers 20,000 2,000,000 200,000,000 Orders 1,000/month 100,000/month 10,000,000/month Products 10,000 100,000 1,000,000

Methodology

A DS2 load was generated within a virtual machine. The load was gradually increased by adding extra DS2 threads. As a result, the database performance should linearly scale up until it would reach a saturation point. Any further added load would only lead to a marginal performance gain or even a performance degradation. Each DS2 run would start some I/O to warm up the SQL Server database cache for one minute. Afterwards, DS2 statistics were reset and the run would continue for ten minutes. At the end of each run, an average Orders Per Minute (OPM) was collected. The average OPM reflects sustained database performance during the period rather than the peak performance. Additionally, two other performance metrics on the database server, average CPU utilization and average disk throughput, were collected using Microsoft Perfmon. Between DS2 runs, the database was cleaned up to reset the database tables back to default. The SQL Server database service was restarted so the database cache was reallocated to eliminate the dirty cache impact from run to run. SQL Server performs extensive I/O caching. To examine caching effect on performance, we increased the database servers virtual machine memory from 1GB to 2GB and then to 4GB. The DS2 database was dropped and recreated before DS2 runs were started to guarantee a clean environment for the next round of testing.

Performance

Figure 4 summarizes the scaling of database performance when storage was served using iSCSI. The chart illustrates that as new DS2 threads were added to increase the database load, the database performance scaled up nearly linearly until it reached 4242 OPM, 5275 OPM, and 5277 OPM, respectively, depending on the RAM size on the database server. At this point, the Iomega storage device was saturated and could not yield any higher database performance.

6 of 8

DVDStorePerformance SQLServer

3 2.1.0.5 0

OrdersPerMinute(OPM)
GuestAverageCPUUtilization #ofThreads 1GBVM 2GBVM 4GBVM

DiskThroughput(MB/s)

Figure 3 DVD Store performance iSCSI
Table 3. Performance numbers database server with 1GB RAM

# of threads 9

OPM 4115
Guest Average CPU (%) 25 26.1 29.1 28.5 35.9 36.9 37.2 37.7 38.3
Disk Throughput (MB/s) 1.9 2.31 2.47 2.48 2.6 2.79 2.71 2.75 2.73

7 of 8

Table 4. Performance numbers database server with 2GB RAM

OPM 4952

Guest Average CPU (%) 26.6 39.8 40.5 41.9 38.7 43.58.3 32.7
Disk Throughput (MB/s) 2.08 1.9 2.26 1.2.02 1.68 2.34 2.27

Table 5. Performance numbers database server with 4GB RAM

OPM 4998

Guest Average CPU (%) 27.1 30.1 24.1 46.5 42.57.5 43.6 25.9
Disk Throughput (MB/s) 1.86 1.96 2.17 1.79 2.06 1.91 1.82 1.96 1.79

Conclusion

The Iomega StorCenter ix4-200r is a unified NAS and IP SAN storage specifically designed to meet SMB business requirements. The storage device contains four SATA disks up to 1TB each, making it extremely affordable to meet the storage capacity needs of many SMBs and remote work groups. Although SATA drive is most suitable for applications that create a sequential I/O pattern, it can be used to serve applications that generate random I/Os. The key is to size and design the solution appropriately based on the underlying storage performance characteristics. The exercise described in this white paper proves that the Iomega ix4-200r is an ideal storage platform for SMBs or remote work groups that require the ability to deploy a small database application in their virtualized infrastructure. The storage device is easy to manage, very reliable, feature-rich, and can provide sufficient performance to small database applications that are deployed in an environment as simulated in this paper.
2009 Iomega Corporation. All rights reserved. Iomega, StorCenter, and the stylized i logo are either registered trademarks or trademarks of Iomega Corporation in the United States and/or other countries. EMC, Lifeline, and Retrospect are registered trademarks of EMC Corporation in the U.S. and/or other countries. Certain other product names, brand names and company names may be trademarks or designations or their respective owners. Iomega's specific customer support policies (including fees for services) and procedures change as technology and market conditions dictate. Product capacities are specified in gigabytes (GB), where 1GB = 1,000,000,000 Bytes. To obtain information about Iomegas current policies please visit Iomega at www.iomega.com DEUG-0909-01