Hi everybody, Last year I implemented Exchange 2010 in a small, high-volume, non-redunant environment. This year I am at a new employer and am working on getting ready to implement 2010 in a large, low-volume, highly-available environment. I'm working on getting the disk configured on the mailbox servers and I've started using the JetStress tool and I'm having a tough time with it...or at least interpreting the results. I am running one server with a RAID 10 array containing the OS and JetStress install. I then have five 1TB disks mounted as mountpoints to a volume in the RAID 10 array (E:\MB001, E:\MB002, et al). As an aside, for now, we will be "Piloting" Exchange 2010 with two mailbox servers with five of these disks, but eventually will have four of them with 20 disks each. Configuring the tool is simple enough, but I think I'm merely confused at the "Thread Count" section as it keeps failing. I read the Appendix A in the JetStress Field Guide, but it isn't making a lot of sense to me. E.g. what is "Storage max"; the maximum amount of disk space for the server, per disk, or the largest I want a database to be? Where is the "average of all total thread count values" of 135 come from? Can anyone give me a few pointers or demystify the guide a bit? I'd be ever-so-grateful! I have used the storage calculator, and while it seems easier to use than this, maybe I'm simply not getting all of the information out of it properly. Thanks in advance!

Basically you must run PerfMon to measure the amount of IOPS being thrown by Jetstress and adjust the thread count up or down to achieve the IOPS you want.Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

Thanks for the reply, Ed. So I should be running PerfMon while I am running the test? Which I/O-related counters should I be watching? How do I know how many IOPS I want? I know more is better, but I'm curious as to how to quantify that. Is that from the bottom section of "Role Requirements" tab of the Storage calculator?

Basically you design your I/O subsystem to handle the number of IOPS you want it to handle. If you use the Mailbox Role Calculator that Microsoft provides, it'll give you the required number of IOPS. What you want to make sure is that Jetstress is throwing the correct number of log reads, log writes, database reads and database writes to simulate your anticipated load.Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

Where do I get it from on the Calculator though? At the bottom of the "Role Requirements" tab there is a "Host IO and Throughput Requirements" section where it says "Total Database Required IOPS" is 5/database. So is that the required IOPS that I need? How do I calculate the necessary thread-count from there? I've tried as high as 20 and as low as 4 and it fails each time.

There's a place where you'll see the IOPS for the server. If you're looking at 5 IOPS per database, you can almost put that database on a floppy disk. That is not exactly a high number. Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

That's why I am confused. I will eventually be be looking at 2500 users (5GB mailbox limit each), 4 servers (three hosting 1/3 of the live mailboxes and the fourth only hosting copies) each with 24 1TB disks. I don't understand why I'm getting "Total Database Required IOPS" of 5/database, 113/server, and 450/DAG...unless I am simply entering the data into the calculator incorrectly... Keep in mind that for my pilot I will only be doing two servers with five disks each and 100 total users, which changed my "Total Database Required IOPS" to 6. From what I can view on the Performance HTML reports it is my I/O Database Reads Average Latency which is my problem, as they are all over 20ms. EDIT: As for my disk configuration, all disks (except the system volume) are 7200RPM 2.5" SAS 6Gb disks on an IBM ServeRAID M5015. This controller doesn't support JBOD so I created five 1-disk RAID 0 arrays with a 256K stripe and I formatted each of the disks in Windows as NTFS with 64K allocation unit size.

Okay, so I would have answered my own question if I fully RTFM. :p So the 133/server is what I needed, and for whatever reason I had a mental block that dropping the number of threads was a bad thing to do. So the total IOPS I am getting is around 440, so I think I am in the clear. ;) Thanks again, Ed. :)

Happy to have helped!Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."