High CPU utilization is not (by itself) a problem. OS/MVT (and all of its 
successors, OS/MVS,  MVS/ESA, OS/390. z/OS..) have been designed (and 
redesigned)
to run at 100% utilization for months and years at a time (I know, I have been 
there).

That being said, most things will run (FSVO) better when there is a little 
excess capacity.

There are so many variables that can contribute to elapsed time that I can't 
even begin to provide any sort of meaningful discussion about any of them.
A partial list of things to look at:
PR/SM overhead (too many logical processors defined)
Workload manager policy
Device contention
DFHSM (or equivalent) recall
Virtual Tape recall
Human factors (physical tape mounting)
Job Scheduling
Job Contention
And many more...

One key to look at is the RMF Mon I CPU report. The key number is the "IN 
READY" queue.
The ideal number is (number of logical processors * 1). That is, one job is 
always ready to execute.
IN READY < 1 indicates excess CPU capacity
IN READY > 1 indicates excessive workload (FSVO excessive). There is no Rule of 
Thumb here. Whatever you can tolerate is a good number

As other have said, are SLA's being missed? 

HTH,

<snip>
In environment we are using 2818-X02 box with around 1100 MIPS and this boxes 
is utilized by couple of LPARS(SYSP & SYST). One is a PROD and other is Play 
LPAR and almost 90% of weight is contributed to PROD SYSP system.
We have a scenario of High CPU utilized only during our batch window(4hours) 
and few days back the system was recycled due to the high CPU utilized and we 
feel that this may create impact during Peak season in November, December.
Also we got a report from our batch team and most of batch  jobs are using more 
CPU time with very high elapsed time and below is for your ref.

JOB NAME        CPU TIME        ELAPSED TIME
Job A           29:22.1         01.48.26.24
Job B                     19:12.3               02.44.26.24

We are position to bring down the CPU Utilization time and make the system 
healthy and we are trying to split the Batch jobs and run on different time.
I would like to get some clarity on the Elapsed time, whether this elapsed will 
make use of the CPU and will it be the cause of high CPU utilized during batch 
window?
</snip>
 

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to