Re: HP IDS 7.3 Performance problems.

From: "Kimberly Dicken" <infosys@ecke.com> > >Our HP Guy(tm) seems to want to do nothing except a) throw RAM at the >problem, and/or b) throw disks at the problem. I'm not completely confident >that the problem is solely I/O as the new machine has more disks, and >better controllers. Backups are 15 minutes. Can you supply us with output from onstat -R, onstat -F, onstat -p, onstat -l and onstat -g seg? >Informix is currently using 30000 buffers, and a SHMVIRTSIZE of 60000. It 30000 buffers seems a bit low. >is my understanding that HPUX doesn't like having more than 2 or 3 shared >memory segments, otherwise its performance goes way down. Now, I don't know >if this is a system limitation, or process limitation as to how many >segments the process is attached to. According to ipcs, informix owns 2, >and the system has 7 total segments. That's interesting can you send the ipcs output as well? >I've only got one CPU configured. Checkpoints are running about every five >minutes, and hit for roughly 3-5 seconds (though there have been some >spikes of 10-20 seconds). Try adding another one or even two CPUVPS. >----------8<--------------------8<--------------8<-------------- >PHYSDBS rootdbs # Location (dbspace) of physical log >PHYSFILE 1000 # Physical log file size (Kbytes) >LOGFILES 10 # Number of logical log files >LOGSIZE 20000 # Logical log size (Kbytes) You should move logical and physical logs out to separate spindles and you should allocate at least 3 temp dbspaces. >TBLSPACE_STATS 1 Disabling this might help a little. >LTAPEDEV /dev/null # Log tape device path ARG!!! NOOOOOOOOOOOO!!!! >NETTYPE soctcp,10,10,CPU # Configure poll thread(s) for nettype Bad move: try soctcp,1,100,NET >RESIDENT 0 # Forced residency flag (Yes = 1, No = 0) Set to 1. >MULTIPROCESSOR 0 # 0 for single-processor, 1 for>multi-processor Try 1. >NUMCPUVPS 1 # Number of user (cpu) vps And 2 or 3 >SINGLE_CPU_VP 1 # If non-zero, limit number of cpu vps to>one And 0. >LOCKS 50000 # Maximum number of locks Low. Try x2 or x3. >BUFFERS 30000 # Maximum number of shared buffers Try x2 or x3. >NUMAIOVPS # Number of IO vps Is KAIO enabled? (onstat -g ath | grep kio) If not, try setting this to at least 9 or 10...avoid powers of 2. >PHYSBUFF 32 # Physical log buffer size (Kbytes) >LOGBUFF 32 # Logical log buffer size (Kbytes) Check onstat -l to see if you can tweak these. >LOGSMAX 20 # Maximum number of logical log files >CLEANERS 4 # Number of buffer cleaner processes A bit low. Try 11 or 12. >SHMVIRTSIZE 60000 # initial virtual shared memory segment>size See how many type V segments are reported by onstat -g seg while the system is under stress. If you see more than 1, increase this. >LRUS 4 # Number of LRU queues Try 9 or 10. >LRU_MAX_DIRTY 7 # LRU percent dirty begin cleaning limit >LRU_MIN_DIRTY 5 # LRU percent dirty end cleaning limit Is this and OLTP system? If so, try 2 and 1, if not, try 80 and 70. >RA_PAGES # Number of pages to attempt to read ahead >RA_THRESHOLD # Number of pages left before next group You might be able to tweek these, try 16 and 12. >DBSPACETEMP tempdbs # Default temp dbspaces You should have at least 3. >OPTCOMPIND 2 # To hint the optimizer Try 0. HTH. ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com