Re: CPU usage
Graeme responds to Eric w/: > >> Recently another group joined us on our development platform and > >> brought with them an Ingres engine. Performance on the box > >> immediately degraded. Just now I looked out there and discovered the > >> Ingres engine had been up for 36 days - in that time it had racked up > >> 1448:40 of cpu time! Comparatively our FOUR Online engines have been > >> up for 115 days and have racked up a total of :58 - now I don't know > >> what they are doing up there but this seems like quite a disparity. > > >Much of the real work done by OnLine is done by the sqlturbo processes. > >They do the parsing, optimizing, disk reads and writes. The Tbinit > >processes simply coordinate this effort among all the other sqlturbo's > >so that no two processes duplicate work. > > Remember that the usercpu and syscpu counts in the tbstat -p profile > reputedly reflect the sum of the cpu resources used by all user > processes (which I have always interpreted to mean sqlturbos). > > What's the total cpu for all four sets of daemons plus all four profiles > Jack? > This is the note I was about to send before I got wrapped up in something else. I think it answers part of Graeme's daemon question. < Alan Points out: (joined by Eric soon thereafter) < < > My immediate impression is that Informix is not accounting accurately. < > < > Only 58 minutes in 115 days! You gotta be kidding! Wait a minute, < > that must be seconds, since 1448/36 = 40+. So Ingres uses 40+ MINUTES < > of CPU a day. Not too bad. < > < > Either something rolled over and reset Informix's minutes to zero, or < > Informix engines keep spawning subtasks which rack up the time and then < > go away, taking their bill with them. < > < < I went out and checked - We've got, on production: < < The preeeemo engine (up 23 days) has 30 minutes for its tbinit < (up 40 days) 61 minutes for its seqlexecd < < (Interesting - the engine was rebooted w/o restarting the sqlexecd - phone < isn't ringing - it must still work) < < (~100 users) 26:42 for sqlturbos, 13:22 of which < (1hr 45 min) is the transaction processor < < no front ends - they are on terminal servers < < Looking back at the development machine there are a bunch of other ingres < things running around - I don't know enough to know what they are, just enough < to know that they belong to that group. They aren't chewing up too much time < (just the iislave and iidbms procs do that) but there is no way they are < doing the same amount of work as the production machine. < < So, as Eric and Alan pointed out Informix is 'hiding' the bill in the < sqlturbos. But is the bill as high? Probably should stow this until I < know more about Ingres. < < Maybe I should do a panty raid on comp.databases.ingres. < < j. But now that Graeme mentioned tpstat -p I went out and checked that. "I see!" said the blind man as he picked up his hammer and saw. That looks more like: user sys time up (minutes) (minutes) production 14972 811 5.21 (5 days 21 hours) 15982 2325 5.21 1308 517 1.16 145 56 1.22 development 565 60 1.22 2996 301 122.21 23541 1651 97.22 946 12 97.22 I should properly relate this to the amount of work done, but my basic question has been answered. Yes, Informix is 'hiding' the bill from 'ps' - but they are hiding it in plain view - 'tbstat -p'. No the Ingres numbers are not inordinately high - and I shouldn't go beat on them about it. Thank you all, j. _____________________________________________________________________________ Jack Parker - Contractor | Hewlett Packard, BSMC Boise, Idaho, USA| If you keep staring at it like that, jparker@hpbs2561.boi.hp.com | your nose is going to grow (208) 396-5388 (W) (208) 384-1623 (H) | into the bark. _____________________________________________________________________________