03-19-2014 05:45 AM
We use for our SAP servers HPUX IA 11.31 with Oracle 11.2
During a load proces (much Data) the performance is realy bad, although in GPM some disks are at 100%.
There are 2 Luns of 800MB, mirrored over 2 F400 3Par storage stustems
The queue depth of both luns is 8
Is this value correct or should I increase it?
Groeten, Alfons velthof
03-20-2014 11:37 AM - edited 03-25-2014 10:03 AM
Can't see the picture. You'll have to attach the picture.
But if LUN has a very heavy load, it is because application(s) are using that particular LUN.
The first question to ask is:
--Is that normal for this application or is this something new?
A database search for fields without indexes is typically a 100% load for serial reads. A correctly designed database will eliminate this load.
You can run sar -d 1 5
and see the load on this LUN. If the AVSERV time is high for this LUN, it may indicate a LUN layout problem or very random access, possibly due to two or more disparate areas on the disk with heavy access. Possibly moving some directories to other volume groups may help.
03-31-2014 11:43 PM
Until the near past we always created at least 4 lun's in a Volume group, because it would increase the io performance.
Is this still valid as we do have very fast Disk systems these days?
Best regards, Alfons
04-01-2014 01:12 AM
>Is this still valid as we do have very fast Disk systems these days?
Well, since 3PAR's VV are stripped over multiple PDs, not so much.
But multiple LUNs may be helpful if each is connected to a different node.