Load Simulator - Compaq 1850R - ProLiant - 128 MB RAM Performance Notes

             
Second, keep in mind that benchmarks do not account for issues such as backup and disaster
recovery or information store maintenance sizing. Whatever the issue, care must be taken when
interpreting benchmarks to ensure that they represent useful information for your Exchange
Server deployment and are based on valid simulation methodologies.
While it is significant that the ProLiant 1850R server can successfully scale to 8,000 medium
MAPI e-mail users in a single site and single server benchmark exercise, Compaq does not
recommend deploying 8,000 users on this configuration. Compaq recommends careful
evaluation of all issues involved in real-world Exchange Server deployments such as
management, administration, and disaster recovery.

Load Simulator

The main tool used in generating the workload for this scalability demonstration was the
Microsoft Exchange Server Load Simulation utility called Load Simulator. Load Simulator is a
tool for simulating a client user load on an Exchange Server. Its purpose is to enable a single
Windows NT machine called a LoadSim client to simulate multiple Microsoft Exchange client
users.
The operation of Load Simulator users is governed by a Load Simulator profile. This profile
controls factors such as how long a Load Simulator "day" is, how many e-mail messages to send
in a day's time, how many times to open and read e-mail, whether to use distribution lists,
whether to use public folders, etc.
Load Simulator creates a highly accurate simulation of reality. It mimics the full Microsoft
Exchange Client in many respects. First, it uses .MSG files, the same format used by the
Exchange Client. This guarantees that messages generated by Load Simulator have the same
properties as those sent by real users of the Exchange Client. Second, Load Simulator uses the
same MAPI remote procedure call (RPC) semantics as those used by the Client. Third, Load
Simulator registers MAPI change notifications in the same manner as they are registered by the
Client. Finally, Load Simulator even emulates the Microsoft Exchange Client list box cache,
which the Client uses for folder and message panes in the viewer when a user browses and selects
messages on the server.
Several steps are necessary to perform a successful simulation. The Load Simulator setup and
initialization process comes first. Load Simulator creates the test topology by generating the user
directory entries. Next, the test store is initialized and populated with the test messages and
folder items. The tests are typically run for up to 8 hours depending upon the user load simulated
and amount of time required in reaching a steady state for measurement purposes. During a test
run, users log on to the Exchange Server and begin processing various messaging tasks. Task
response time data is logged to the LSPERF.LOG file and client messages and error logging is
stored in the LOADSIM.OUT file. To produce test scores, the LSLOG utility is used to parse the
LSPERF.LOG file and calculate the response time score. By default, 95th- and 50th-percentile
response time scores are calculated. Ninety-fifth-percentile response time scores for the
MAPI/RPC protocol should be less than 1,000 ms, according to Microsoft. Also, the Exchange
Server IS Send Queue and the MTA Work Queue (other message and delivery queues should also
be considered, depending on the protocol) must consistently return to near zero during the steady-
state period for which test measurements are taken. Queues that continue to grow and fail to
return to near zero indicate that the server is not sustaining the required workload. There should
be no errors logged by the LoadSim clients during the test. When these conditions are met, a
successful test run has been completed. For more information on LoadSim Medium canonical
profiles, please refer to the LoadSim documentation at:
http://www.microsoft.com/exchange/library/loadsim55x86.exe.
ECG1080898