Opinions on usercpu & syscpu in onstat -p

Just wanted to get some opinions on the usercpu and syscpu meaning in the onstat -p output. I can't remember if I read in Lumbley's guide or not,but if usercpu exceeds the amount of lapsed time this indicates the system could benefit from more physical cpus. Below is an onstat -p output where total elapsed time is 67740 seconds. User cpu more than twice that at 150,951 HP N Class 6 Way 550's 6 gig Physical Mem. HPUX 11 9.40 FC2 IBM Informix Dynamic Server Version 9.40.FC2W4 -- On-Line -- Up 2 days 20:32:50 -- 3918780 Kbytes Profile dskreads pagreads bufreads Êched dskwrits pagwrits bufwrits Êched 58767865 124325277 12739374150 99.54 4592164 31893241 24004499 80.87 isamtot open start read write rewrite delete commit rollbk 7825834138 6479678 83176707 6822458070 178820294 1276025 720792 80860 488 gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs 0 0 0 0 0 0 0 ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes 0 0 0 150951.57 5282.00 143 429 bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans 3446881 2048 12220794081 23 0 512 480281 5784056 ixda-RA idx-RA da-RA RA-pgsused lchwaits 13697305 1729291 18192868 33137079 94700177

This output must be from a multiprocessor system with more that one cpu vp configured. How many physical cpus and cpu vps does it have? What os? With n physical cpus and n cpu vps onstat -p will show a maximum of usercpu + syscpu = (n * elapsed-time) If you are close to that maximum having more physical cpus will probably be an improvement. But there are other diagnostics like ps -ef sar -u or vmstat onstat -g rea onstat -g act onstat -g glo that should be used as well. Another option may be to tune the application and index strategies instead of buying more cpus. Michael Darren_Jaco.... wrote: > Just wanted to get some opinions on the usercpu and syscpu meaning in the > onstat -p output. I can't remember if I read in Lumbley's guide or not,> but if usercpu exceeds the amount of lapsed time this indicates the system > could benefit from more physical cpus. Below is an onstat -p output where > total elapsed time is 67740 seconds. User cpu more than twice that at > 150,951 > > HP N Class 6 Way 550's > 6 gig Physical Mem. > HPUX 11 > 9.40 FC2 > > > IBM Informix Dynamic Server Version 9.40.FC2W4 -- On-Line -- Up 2 days > 20:32:50 -- 3918780 Kbytes> > Profile > dskreads pagreads bufreads Êched dskwrits pagwrits bufwrits Êched > 58767865 124325277 12739374150 99.54 4592164 31893241 24004499 80.87 > > isamtot open start read write rewrite delete commit > rollbk > 7825834138 6479678 83176707 6822458070 178820294 1276025 720792 80860 > 488 > > gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs > 0 0 0 0 0 0 0 > > ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes > 0 0 0 150951.57 5282.00 143 429 > > bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans > 3446881 2048 12220794081 23 0 512 480281 5784056 > > ixda-RA idx-RA da-RA RA-pgsused lchwaits > 13697305 1729291 18192868 33137079 94700177 > > > >