RE: Job run times occasionally seriously spike. How do I start to figure this out?

Rob Berendt <rob@xxxxxxxxx> wrote on Tue, 22 Nov 2016 14:20:03 GMT:

Is there a "getting started" on some of these?

There are classes available from Lab Services which can educate on how
to use Performance Data Investigator to track down specific performance
issues. Dawn May's iCan blog and developerWorks have pointers to
presentations and articles if you want to go the self study route.

<snip>
Are disk response times equivalent for all runs?

I like the new Disk Reads and Writes perspective. Information on how to get to it and the information it contains is at
developerWorks
https://www.ibm.com/developerworks/community/wikis/home?lang=en#!/wiki/IBM%20i%20Technology%20Updates/page/Disk%20Reads%20and%20Writes%20Detail

Are I/O per second (IOPS) equivalent for all runs?

What about the hosting LPAR? CPU, disk response time, etc.
equivalent?

Assuming the hosting LPAR and the client LPAR are same time, you can
pull up the same information on both in different browser windows to
do comparisons.

Lastly, even with your comments about PMRs, supportline will help you
in tracking down a performance anomaly like this. They'll ask for collection
services data to make comparisons and may ask you to set up some job watcher
monitors to submit so they can look deeper into what's going on during a
short run versus a long run. Job watcher takes more skill to work with
compared to collection services, but can be invaluable in finding things
like this.