I have read all of the documentation from technet and various other sources. Yet still, I am having a hard time understanding excatly what the "Thread Count" is when you are running JetStress 2007 to test Exchange 2007 disk configuration. For example, if I configure a test providing a specific number of mailboxes, iops/mailbox and size of the mailboxes...where does thread count come in? I really don't understand what a thread equates to and how it has an affect on the test other than that it increases the load. I want to perform accurate tests regarding projected usage on my Exchange 2007 environment and a better understanding of thread count and how it is used in JetStress will certainly help.

I configure a test providing a specific number of mailboxes, iops/mailbox and size of the mailboxes The number of mailboxes is not actually the parameter of JetStress "Performance" test. It's just one of factors affecting the summary IOPS. First, you have to count the expected IOPS wich you need fora disk subsystem of your server, in dependance of circumstances (enumerated by you). Please reread Optimizing Storage for Microsoft Exchange Server 2003 (page 18 and so on) for a better understanding. I really don't understand what a thread equates to and how it has an affect on the test other than that it increases the load. Next, you're sorting out a necessary number of threads to find one, providing the load approximately equal to estimated IOPS. And at last, you run the "Performance" test with a found threads count for a some significant time. Aftertest's finish, you estimate results of it by server working parameters during it. I.e. latencies, memory pooland so on, and it's comparison with border values. That's all, in a brief.

A kind of surreptitious advertising: http://www.amazon.com/Designing-Storage-Exchange-Digital-Technologies/dp/1555583083 page 438. For the first test run it is recommended to let Jetstress automatically tune the number of threads. For subsequent tests you should disable automatic tuning and manually specify the thread count. This allows you to have comparable test conditions. Manually specifying the thread count is especially important if you run Jetstress in parallel on multiple servers. If you use automatic tuning then one system can still be in the tuning phase while other systems already execute the test. This would lead to inconsistent load on the subsystem and the test result would be different for each test run. The section Using Thread Count in the online documentation provide a table that recommends number of threads based on the expected log disk write I/O and disk transfer rate on the database volume. Please be aware when you increase the thread count then this increases the thread count per storage group. The change to the total workload for the server is significant different between a server with 1 storage group and a server with 20 storage groups.