RE: IDS 7.30 CPU Usage

Comments follow. >===== Original Message From "Jay Turner" <tjay@edcpub.com> ===== >Back in November we upgraded both our OS and Informix engine. From AIX >3.2.5 to AIX 4.3.2 and SE 5.03 UC1 to IDS 7.30 UC5. We are still using the >same system (RISC 6000 59H 768 meg. memory) and programs (4gl programs >recompiled with I4GL 7.20 UD6). We were told to expect a huge performance >increase, well it hasn't happened. Actually as a whole it is slower, the >problem appears to be that when we get a moderate amount of activity the >online process starts taking up all the CPU resources making the whole >system run slower. After studying several performance related articles, >books, newsgroup posts and inquiring with other companies who have done >something similar, we have come up with no solution. Does anyone have any >suggestions? We really need to get this figured out. We are at a point >where we need any help we can get and would be willing to pay someone to >come to our location or help us remotely. > >Thanks for any help, >Jay Turner > >Here is our onconfig file and a onstat profile. > <SNIP> ># Physical Log Configuration > >PHYSDBS rootdbs # Location (dbspace) of physical log >PHYSFILE 12000 # Physical log file size (Kbytes) If possible put this in a separate dbspace and disk from rootdbs, your logical logs and active dbspaces. It is a log which does a lot of I/O. ># Logical Log Configuration > >LOGFILES 6 # Number of logical log files >LOGSIZE 4000 # Logical log size (Kbytes)> If possible put this in a separate dbspace and disk from rootdbs, your physical logs and active dbspaces. It is a log which does a lot of I/O. <SNIP> >TBLSPACE_STATS 1This parameter causes overhead. Unless you are using these statistics to monitor how your programs access data. Set it to 0 to turn it off. <SNIP> >LOCKS 30000 # Maximum number of locks >BUFFERS 24000 # Maximum number of shared buffers >NUMAIOVPS 2 # Number of IO vpsAre you using KAIO? If not, how many disks do you have, NUMAIOVPS might need to be higher. >PHYSBUFF 32 # Physical log buffer size (Kbytes) >LOGBUFF 32 # Logical log buffer size (Kbytes)>LOGSMAX 6 # Maximum number of logical log files >CLEANERS 24 # Number of buffer cleaner processes >SHMBASE 0x30000000 # Shared memory base address >SHMVIRTSIZE 75000 # initial virtual shared memory segment size >SHMADD 16000 # Size of new shared memory segments >(Kbytes) >SHMTOTAL 0 # Total shared memory (Kbytes). 0=>unlimited >CKPTINTVL 600 # Check point interval (in sec)Are your checkpoints long. look in your online log to find out. If so, lower your CKPTINTVL down to 300 and your LRUMIN/MAX DIRTY to 5/10 respectively. >LRUS 24 # Number of LRU queues >LRU_MAX_DIRTY 60 # LRU percent dirty begin cleaning limit >LRU_MIN_DIRTY 50 # LRU percent dirty end cleaning limit>LBU_PRESERVE 0 # Preserve last log for log backup It wouldnt hurt to set this to 1 if you go to more logical logs of a smaller logical log size. It will save your last logical log for administrative tasks. This way you have more options if you end up running out of logical log space. <SNIP> > > >Informix Dynamic Server Version 7.30.UC5 -- On-Line -- Up 6 days >22:42:49 -- 175456 Kbytes >Profile >dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached >17595855 6484572 514902906 96.58 3440922 4189476 23870640 85.59 > >isamtot open start read write rewrite delete commit >rollbk >329530188 22069074 42814341 135475765 2744602 5447069 1391585 125630 0 > >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 182881.09 6356.53 1573 3500 > >bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans >1319738 850 177855623 0 0 2878 376052 586287 > >ixda-RA idx-RA da-RA RA-pgsused lchwaits >3609922 53840 8555932 12213105 1607 Are you doing UPDATE STATISTICS. If so, are you following informix's guidelines for it. Somone else mentioned you might be running into a bug having to do with indexes, doing an onstat -P | tail -20 and looking at the percent of B-tree buffers seeing if it is extremely high will show if you are running into this bug Hope this helps Will ------------------------------------------------------------ This e-mail has been sent to you courtesy of OperaMail, as a free service from Opera Software, makers of the award-winning Web Browser, Opera. Visit us at http://www.opera.com/ or our portal at: http://www.myopera.com/ Your free e-mail account is waiting at: http://www.operamail.com/ ------------------------------------------------------------