Re: qtemp performance oddity.

These 8 17s just replaced 8 7200 RPM 8GB Units (Remember the big black 1.6" think drives?!) and the RAID card from 2741 to 2778.

After the upgrade MANY batch jobs went from 15 seconds to 10 to 20 minutes and signoff for interactive users went from near instant to more than 2 minutes.

Yes performance reports do show the 8G units at almost exactly 1/2 the total ops of the 17s. HOWEVER with one exception of a .018 all the 6717s also show average response times of .020 just like the 17s.

Do you think we have a RAID Card issue here?

- Larry "DrFranken" Bolhuis

www.frankeni.com
www.iDevCloud.com
www.iInTheCloud.com

On 2/28/2013 11:57 AM, Sue Baker wrote:

On 2/28/2013 10:28 AM, DrFranken wrote:

Disk response times aren't awesome but they are 10K SCSI on an old 720.

But those old 10K SCSI should still be able to average around 12-15ms
not 30+ ms!

With your other post saying they have a mix of 8 and 17's on the system,
I'm willing to bet 2/3s of the QTEMP data is going to the 17G drives and
1/3 to the 8G drives based on percent full. So, I would expect system
report and/or component report disk sections to show the 17G drives are
2x the 8G drives from an IOPS perspective.

Waits are probably much longer on the 17G drives than on the 8G drives
because even with low utilization, if the drive is busy servicing a
request, the op has to wait for the drive.

You could potentially use TRCASPBAL followed by STRASPBAL *USAGE to move
data around on disk to influence the behavior. This would have to be
done on a regular basis .... IE trace every day and balance on the
weekend, or trace every day and balance every day.

Otherwise, replace those 8G drives with 17G drives.