Performance Issues

[Guest.Visitor]

Also review the System Values for QTOTJOB & QACTJOB to ensure these are both substantial enough to hold the number of jobs within the system. If these values are too small, the performance degrades as the system initializes for increased number of Active Jobs and/or Total Jobs. Currently, we have our system QACTJOB set to 1000 and QTOTJOB set to 15000. These values as shipped with the system are set to QACTJOB (20) and QTOTJOB (30).

[David Abramowitz]

You could test out performance after changing storage pools from *FIXED to *CALC. This turns cache on. It does not guarantee better performance, but better performance may be a result. Try it and see. There are other generalities which may result in better performance. Production queries can take a long time to run, and can slow down other jobs. Change them to HLLs. Batch job streams that run sequentially often have several jobs within the stream that may run concurrently. Finding them can knock off significant time off of the run. The smaller the interactive job, the less memeory is used. Try to avoid clutter in screen design. Check the number of LFs on every PF. The general rule is anything more than seven permanent LFs may affect performance. See what can be turned into an OPNQRYF, or what may be obsolete. Seek time is performance costly. The more DASD that is occupied (over 82%) the slower your jobs will run. Identify obsolete objects, back them up, and then get rid of them. Please keep in mind that the above statements are generalities, and not hard and fast rules. Dave

[brunfish@newnorth.net]

I do not understand why you changed this from 2 to 1. This is telling the OS to do less performace adjusting and would be a negetive move..

[tim.shafar@fiserv.com]

I was told by someone that adjusting on the fly is a bad move. It's like playing catch up to your systems use. Because the adjustment comes after the fact. Lets say I have several large batch jobs that are submitted. (They are the exception rather than the rull.) Now I just lost Interactive resources because the of the adjustment. Plus the batch will not be hit that hard the rest of the day. Now I have hurt my interactive performance based on an adjustment that came after the fact of when I needed it. that is what i was told....I don't really know. Is there better information on this?

[Guest.Visitor]

My system has a very high number of "total jobs." The number is well over the QTOTJOB system value maximum parm. My question is this..... What if I changed the system value to allow spool files to detach from ended jobs in order to decrease the number of total jobs on my system? System value QSPLFACN set to *DETACH. Would this help with memory allocation with the large total job number? Scott

[tim.shafar@fiserv.com]

We have a developement machine that people are starting to complain about performance issues with. I have checked joblogs from when they said things were good and now and I'm not seeing much of a change in CPU time. I have found more jobs running on the system too. So I'm not really concerned. However, it did make me realise that I'm not very familar with checking on performance issues. Thus my question. What are some of the things I should watch for or check if there is a concern for posssible performance problems? I have the checked the System val QPFRADJ and it was set to 2. I changed this 1. I have watched the values in DSPSYSSTS I know this seams week but what else can I check?

[Guest.Visitor]

David's suggestion is a great place to start: You could test out performance after changing storage pools from *FIXED to *CALC. This turns cache on. It does not guarantee better performance, but better performance may be a result. Try it and see. You cannot change your machine pool ro anything other than *Fixed. You may try to set your QPFRADJ to 3. If you're operating at high storage capacity, you'll see noticible performance problems. What % are you operating at? When you notice performance problems, WRKACTJOB/place your curser on CPU%, press F16 to verify you don't have jobs taking abnormal amounts of your CPU. From here press F11 and place your curser on AuxIO and press F16 to verify you don't have jobs processing abnormal numbers of IO's. Make sure you allow enough elapsed time to get accurate readings.