Performance Monitor showing a stressed Server!

[FragglePete]

Always had my suspicions one of the servers was running a bit slow, always put it down to it's quite old and had earmarked it for replacement, but had a look at the performance monitor and attached is what I saw! Compared to the other servers, this performance trace is a little busy to say the least.

The server is our main file share for peoples home drives (attached external storage device via a 4Gb FC Card), Second DC and main Print Server running Alt-N excellent PCounter software.

The server still seems to work ok, print services and file sharing all seem to be working ok but have had the odd issue with it.

Where to start, no particular service on the server is taking up any large CPU time, seems to be disk access problems more than anything but as I said, files are coming in from the attached Storage device on the FC Card.

Thanks in advance

Pete

[wlaidlaw]

I would break down the Avg Disk Queue counter to be per volume rather than a total of all volumes.

This will show you where the queue is forming. You then know where to spend your money!

I have converted some RAID5 arrays to RAID1 and RAID10 (dependent on volume size) and the performance increase has been incredible. Don't assume that just because you have good kit (SAS drives for example) you can ignore the limitations of the equipment. Schools can (and do) regularly generate enough traffic to exceed these.

[FragglePete]

I would break down the Avg Disk Queue counter to be per volume rather than a total of all volumes.

This will show you where the queue is forming. You then know where to spend your money!

I have converted some RAID5 arrays to RAID1 and RAID10 (dependent on volume size) and the performance increase has been incredible. Don't assume that just because you have good kit (SAS drives for example) you can ignore the limitations of the equipment. Schools can (and do) regularly generate enough traffic to exceed these.

Doing that shows the bottle neck is on the E: drive which is on the LSI Fibre Channel HBA Adapter. The adapter is a 4Gb FC device and the network card on the server is running at 1Gb; could be this bottle neck causing these high readings.

Pete

[wlaidlaw]

Probably the easiest way to find out is as follows:
Prepare another temporary volume, on a (relatively) unused (in terms of access) set of disk(s).
Whilst you are seeing the high queue to your E: volume, create a lot of I/O to your temp volume (for example copy a few big files to here).
Watch the disk queue on E: - If it starts to rise dramatically, stop your test I/O - your issue is likely to be communication to your physical disks. This could be the FC comms or it could be the controller backplane itself.
If the queue stays roughly the same you can pretty much say that the limiting factor is the disk grouping itself.


Dependant upon your storage system you may be able to get some good stats from that... Often you need to use the command line to retrieve a csv file.

Hope this helps

Will