High Physical Disk Queue on Exchange 2003 mail store
I have a raid 10 array with 10 7200SATA rpm disks. My disk line up size throughout organization hrs is balancing about 100. The following holds true of the arrangement:
- The array has one mail Store with 95 energetic mail boxes. (This is the only point, no logs or system documents)
- Avg Mailbox dimension of ~ 400 MegaBytes
- The array is one huge 1.3 TB dividing that was straightened to the raid red stripe
- The Mail store is ~ 48 GB (for both etm and also stm documents)
- The mail store was simply defragged
- The purchase logs get on an additional array that has a much less than 1 avg disk line up
Does this appear high? If so, does anything appear incorrect with this arrangement? Should I consider a few other counters?
Updates after remarks:
- The array itself appears all right, simply the various other weeks it obtained excellent arise from numerous days of jetstress screening
- There is no pagefile on the array: -)
- There is symantec AV software program. The Highest 2 of IO Reads and also IO Writes from considering Task Manager are conduit.exe (symantec anti spam / infection) and also store.exe. avenue goes to 18 million reviews and also 25 million creates, store is 144 million reviews and also 9 million creates. Already, given that I have portal web servers, so I am checking into taking AV off of the backend web server
That is instead high, do you have any kind of sort of AV software program? Additionally consider bkslshprocess *bkslshio information operations/sec. That need to inform you if it is store.exe or somehting else that creating the IO. If it is store.exe I would certainly presume that something is scanning thru your mail boxes.
I would certainly first download and install process explorer and also utilize it to see what procedure is in fact creating the high disk IO, is it the details store, or another thing. After that you can take place from there.
It would certainly serve to recognize just how much mail you are are in fact refining, as 95 mail boxes is not a massive quantity, yet if there done in constant customer after that maybe a concern.
Additionally, you have actually possibly currently examined this, yet is your RIAD array ok, if its restoring from an unsuccessful disk that can create substantial disk IO.
Whoa! that is really high. The ordinary line up size need to amount to, or much less than, the variety of physical disk pins, so your equipment is whipping concerning an order of size over where it need to be. This link has the checklist of all the Exchange procedures that create disk I/O, so in addition to Sam & Evan is pointers you need to validate that you do not have uncommon quantities of any one of those tasks (like a mail loop).
You could additionally consider Process Monitor, which logs each accessibility on the disk independently. If something apart from Exchange is utilizing your disks, you'll see it load the checklist of disk accesses really promptly in ProcMon.
You do not state just how much RAM you have, although the specifications you did state show you possibly have a practical quantity. If you are running much less than 2 GB, you could see this sort of actions from the equipment striking the web page documents. See to it the quantity of RAM made use of in Task Manager is much less than the quantity of physical memory mounted in the web server.
That is past "rather high" - - that is exceptionally, mind - blowingly high. I have boxes with over double that variety of mail boxes working on RAID - 5 on old - as - dust 7,200 RPM Ultra160 SCSI drives with a lot lower disk lines up.
Another thing besides Exchange is whipping your disks. I would certainly toss open Perfmon and also chart the "IO Data Operations / sec" in the "Process" object for each and every specific procedure and also see what procedure is creating a lot IO.
The write-up you connected in your comment to Jim B has some great perfmon counters to look at. I'm asking yourself, also, if you've placed a digital memory pagefile onto those disks and also are seeing too much paging.
I do have some uncertainties, after reviewing the write-up and also connected write-ups re: Entourage that you could be having actually some concerns associted with those customers. Expectation Anywhere (also known as RPC over HTTP) isn't mosting likely to create the very same troubles that Entourage will, though - - that is a various point totally (MAPI over HTTP, versus WebDAV which the Entourage customers make use of).
It goes w/o asking, yet are you seeing anything weird in case logs?
Modify after your upgrade:
The complete variety of reviews / creates isn't actually what you are seeking. You are actually seeking the delta in reviews / creates on a per - interval basis. Toss open Perfmon, clear the default counters, and also add some counters for:
- Object: Process - Counter: Data Operations/sec - Instance: conduit.exe
- Object: Process - Counter: Data Operations/sec - Instance: store.exe
You could additionally look at the Microsoft Exchange User Monitor (wonderful little write-up concerning its use readily available at http://www.msexchange.org/tutorials/Microsoft-Exchange-Server-User-Monitor.html). This will not show WebDAV sessions, yet it could offer you some understanding right into what your typical MAPI - based customers are doing.