Large disk queues

[D J Hewetson]

On some of our HPUX Test systems we see large queues building on some of our internal disks, although the utilisation of the disks is reported as low. Typically these disks contain binaries for Tuxedo, AdvantageGen, and Maestro (a job scheduler) which are used by all the testers. Has anyone had the experience of high queues & low utilisation on non-array disks, and found the root cause?

[Steven E. Protter]

Shalom,

Partially,

High queues usually happen when there is more utilization than the disks can handle.

http://www.hpux.ws/system.perf.sh

Its possible with a few LARGE i/o requests to tie up the disks and have the queues build.

The root cause is usually the way the disks are being used and how many processes wish to use them.

I suspect your metric for measuring disk saturation percentage may be flawed or require a patch.

SEP

[Ludovic Derlyn]

Hi,

I have this problem on servers, but i have also a scheduler (Dollar U)

I thinks that number of Io by disks is the cause

regards

L-DERLYN