> I'm having an issue with z/OS 1.4 slowdowns. I've been debugging Java related performance problems for the last 4 years. The problem can be caused by serialization problems between threads. It can be either java implemented spin locks or heavy monitor contention. In order to identify the problem, I usually get a dump of the address space when it is exhibiting the peak CPU usage. Then I run IPCS against it and look at the following: SYSTRACE ASID(X'nn') VERBX LEDATA 'NTH(*)' CTRACE COMP(SYSOMVS) FULL ASID(X'nn) In the SYSTRACE, look to see whether one or more threads from your application are dominating the trace. If nothing jumps out at you, I sometimes run the output to a data set and sort the data by TCB as well as by function to see what is predominate. If I see a TCB(s) that seems to be showing up a lot, I then look at the ledata output to see what the C stack looks like. The CTRACE might help identify failing syscalls or possibly excessive calls. There is also the possibility that you are running into heavy Garbage Collection (GC). I beleive there are JVM options to allow you to report GC statistics. Not sure what the options are. If you run out of options, send me a note.
Don Ault, 8-295-1750, 845-435-1750 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
