How we tested Sun's Storage 7410 Unified Storage System

The test bed for this test consisted of four servers: an HP ML370G5 with dual Xeon 3.0GHz, 6GB RAM and dual gigabit Ethernet; an HP DL360G4p dual Xeon 3.4, 2GB RAM and dual gigabit Ethernet ports; and two SuperMicro-based system each with one Xeon 3.4GHz, 2GB RAM and dual gigabit Ethernet ports. All servers were running Windows 2003 Server. All were connected to a Cisco SLM2048 gigabit Ethernet switch, as were the Sun Storage 7410 controllers. Three of the four 7410 iSCSI ports were aggregated together in an LACP group.

To measure the Sun Storage 7410 Unified Storage System performance, we used IOmeter, a popular I/O subsystem measurement and characterization tool for single and clustered systems that supplies results in both MB per second and IO per second.

There were four sets of workloads running across the servers, each designed to simulate the storage traffic typical of a specific type of server, an Exchange 2003 server, an Exchange 2007 server, a file server and a Web server. Each of the worker processes had an I/O queue of 32 entries, ensuring that the Iometer systems were keeping plenty of I/O ready to go. We believe that this workload at the level we configured far exceeds what a typical server would present.

The Exchange 2003 workload used three workers generating 4-KB I/Os consisting of 67% reads and 33% writes with 95% of those being random and 5% sequential; and one worker generating 64-KB I/Os consisting of 5% reads and 95% writes with 5% of those being random and 95% sequential.

The Exchange 2007 workload used three workers generating 8-KB I/Os consisting of 50% reads and 50% writes with 95% of those being random and 5% sequential; and one worker generating 64-KB I/Os consisting of 5% reads and 95% writes with 5% of those being random and 95% sequential.

The file server workload used two sets of two workers each generating 10% of 4-KB I/Os, 10% of 8-KB I/Os, 40% of 32-KB I/Os and 40% of 64-KB I/Os consisting of 80% reads and 20% writes with 5% of those being random and 95% sequential.

The Web server workload used two sets of two workers each generating 10% 8-KB I/Os, 20% 32-KB I/Os, 20% 64-KB I/Os, 20% 256-KB I/Os and 30% 512-KB I/Os consisting of 95% reads and 5% writes with 5% of those being random and 95% sequential.

We ran each of the workloads for 15 minutes, and repeated each workload test three times, averaging the results to get our performance numbers.

Return to test.