Performance issues after IPL

[Guest.Visitor]

Can any one tell me why after an IPL our AS/400 model 640 on V4R3 appears to go slow. We have noticed a distinct increase in processing time for three to five consecutive days after an IPL. Any Ideas ????

[David Abramowitz]

There are certain batch jobs that have low-numbered priorities. As an example, the SMTP server will usually run at a priority of 35, and gobble up all available memory. You may have chaged these jobs to run at a higher number priority. This will cause all batch jobs to run normally. An IPL will restore priorities to their original settings. Use the WRKACTJOB command, wait a minute and press F5. Then, examine memory usage. If a system job's memory usage is very high, then that is your culprit. Dave

[Guest.Visitor]

David, We have not changed our system job priorities. Our machine has been identified by IBM as processor bound, I am wondering if there is any routine which the system runs (in the background) after an IPL which may be taking longer due to this fact.

[Guest.Visitor]

I wonder if you have QSysVal QPfrAdj set to a value of '1'. This might explain your situation, it could be that the memory allocations for the subsystem(s) were set to "bad" values and the dynamic allocation of memory will not occur. You might also try a WrkSysAct to see if you are being punished by a QCIntJob. Bill

[Guest.Visitor]

Bill, I have checked QSysVal QPfrAdj and it's set to '0' as I thought. We do not run any of the dynamic performance options. However, we are about to begin investigations. Can you tell me what your impression of dynamic performance adjuster is ? I also checked for QCIntJob. Doesn't show on the wrksysact Cheers Lee

[Guest.Visitor]

I have QPfrAdj set to '3' here. The system does a fine enough job here for us. It was on the edge of useless when the feature first came out, but IBM has enhanced the logic behind the memory re-allocations to the point where I am quite happy with it. I may have mis-spelled the Q job you need to look for (actually, I think it is QCFINT1). It is a system job that kicks in on the server models whenever your interactive workload exceeds the defined threashold for your system. This job will kill ALL response times on your system. It does not show up with WrkActJob, it only shows up under WrkSysAct. Bill