Specifications

2
Principled Technologies, Inc.: Dell PowerVault MD3000i high-availability testing on
Windows Server 2008 R2 and Hyper-V
The theoretical peak performance
for our DVD Store workload is
four operations per minute per
user, for both the 200- and 500-
user workloads. With 4 virtual
machines under load, and 12
background virtual machines, all
running simultaneously, the Dell
PowerVault MD3000i solution was
able to sustain near maximum
theoretical SQL Server 2008
operations per minute (OPM): 760
OPM at 200 users and 1,878
OPM at 500 users.
Figure 2 presents the Exchange
Server 2007 latency results of the
Dell PowerVault MD3000i solution
in supporting a 200- and 500-user
count workload in an optimized
system state, where both servers
and all virtual machines are
operational.
Because the LoadGen results are in response time for separate tasks, we created a weighted average to
calculate a final score. To do so, we used a procedure similar to the one that Microsoft Exchange Server 2003
Load Simulator (LoadSim) uses. See Figure 5 for the weighted average we used. For an acceptable score, the
average 95th percentile latency must be lower than 1,000 ms (1 second) in all tasks. With 4 virtual machines
under load, and 12 background virtual machines, all running simultaneously, the Dell PowerVault MD3000i
solution was able to sustain both the 200- and 500-user workload while staying well below the latency threshold:
151.20 ms at 200 users and 517.45 ms at 500 users. This indicates that the Dell PowerVault MD3000i solution
would be able to exceed 500 users running similar workloads in a real-world environment.
In addition to testing the optimized system state, we also measured the performance in a degraded system state.
To simulate the system degradation, we physically unplugged one of the Dell PowerEdge R710 servers and
allowed the system to fail over to the degraded state. The single remaining Dell PowerEdge R710 was then able
to access both its own data and that of the unplugged Dell PowerEdge R710. This remaining Dell PowerEdge
R710 and Dell PowerVault MD3000i solution was able to maintain similarly comparable performance on both
LoadGen and DVD Store workloads, for both a 200- and 500-user count, after degradation. Details of the results,
while the system was in a degraded state, are in the Test results section below.
Dell’s “PowerVault MD3000 and MD3000i Array Tuning Whitepaper” includes configuration and best practices
information, and is available at:
http://www.dell.com/us/en/enterprise/storage/pvaul_md3000i/pd.aspx?refid=pvaul_md3000i&s=biz&cs=555
Workload
LoadGen
Microsoft Exchange Load Generator (LoadGen) is an industry-standard tool for benchmarking an Exchange 2007
Mail Server. LoadGen performs tasks to simulate a standard user generating mail activity. When the workload
finishes, LoadGen reports the response time, which is the number of seconds necessary to complete the task.
LoadGen measures the impact that various clients (MAPI, OWA, IMAP, POP, and SMTP) have on Exchange
servers. This allows users to test how effectively a server running Exchange responds to email loads. The
simulator sends multiple message requests to the server running Exchange, resulting in a mail load.
0
100
200
300
400
500
600
LoadGen latency (ms)
Storage
Dell PowerVault MD3000i
Exchange Server 2007 latency
200 users
500 users
Figure 2: Exchange Server 2007 Performance results of the Dell PowerVault MD3000i
solution in supporting 200- and 500-user count workloads in an optimized system state.
Lower numbers are better.