Specifications

A Principled Technologies test report 7
Hardware upgrades to improve database, SharePoint, Exchange, and
file server performance with the Intel processor-powered Dell
PowerEdge T630
 Mailbox Profile: 250MB mailboxes
 Action Profile: Very heavy
 Client Type: Outlook_500
For more details about LoadGen, see www.microsoft.com/downloads/details.aspx?FamilyId=DDEC1642-F6E3-
4D66-A82F-8D3062C6FA98&displaylang=en.
Collaboration testing
To simulate the traffic of SharePoint Server 2010 users, we used the WSSDW 1.0.0.0 Beta test. This test creates
sample data, populates the server with it, and simulates SharePoint Server users completing everyday tasks.
We used Visual Studio® 2010 to execute the test workload, generating a realistic user workload where each of
the simulated heavy users performed a task every minute. To do this, we set up the test to run 100, 150, and 200 Visual
Studio virtual users with a 30-second think time between each of the various tasks, to simulate 200, 300, and 400 users
for the solutions throughout the comparison and hardware upgrades. The virtual user count here was set to match the
Exchange user count. The test workload then reports throughput in tests per second.
For more details about SharePoint Server, see office.microsoft.com/en-us/sharepoint-server-help/.
File server testing
To simulate file server performance of the solutions, we used the Iometer tool, which measures IOPS and
throughput on both single and clustered systems. Iometer performs I/O operations to stress a system, and then records
the performance of these I/O operations and the system stress they create. We used Iometer 1.1.0 to simulate a file
server workload with varying block sizes and I/O patterns (see Figure 5), with one Iometer worker and 100GB dataset on
both solutions. We used one worker and tuned the number of outstanding I/Os to 1, 2, and 3 to simulate 200, 300, and
400 users, respectively, throughout the comparison and hardware upgrades, each generating between 0.3-0.5 IOPS. For
more details about Iometer, see www.iometer.org.
Block size
Size
(percent)
Reads
(percent)
Random
(percent)
Delay
Burst
Align
Reply
512
10
80
100
0
1
4,096
0
1,024
5
80
100
0
1
4,096
0
2,048
5
80
100
0
1
4,096
0
4,096
60
80
100
0
1
4,096
0
8,192
2
80
100
0
1
4,096
0
16,384
4
80
100
0
1
4,096
0
32,768
4
80
100
0
1
4,096
0
65,536
10
80
100
0
1
4,096
0
Figure 5: Access specification configuration information.