CPU time << elapsed time (trace): Why?

Hello!

Trying to accelerate a long running PL/SQL application, I came to considerate DB physical setup, global parameters and other administrative measures. I used tkprof to evaluate a trace file, where I found that CPU time was considerable smaller that elapsed time. Well, usually, this is due to harddisk activity. The problem ist that we cannot understand this, because we cannot find any reason for this.

Redologs have 2 members spread over different devices, are big enough not to be switched too often, archived log files go to a third device, checkpoints
do not occur that often (because parameters "log_checkpoint_interval", "log_checkpoint_timeout" and "fast_start_io_target" are set accordingly).

Examining v$sysstat and v$system_events does not show any log contention.

How could I find how the rest of the time (not caused by CPU activity) is being spent? I need a more precise method to cope with this. Perhaps the mentioned views do not give any information on wait time of datafiles?? It COULD be that harddisk access to datafiles is the culprit. How can i find out more about that?

Using Oracle 8.1.7 EE / Solaris 7 on an E3500 (4 Ultraparc II CPUs). The devices used for Oracle files are:
/oradata1: 1. members of redologs, other datafiles (mirrored HD)
/oradata2: archived log files, other datafiles (mirrored HD)
/oradata3: 2. members of redologs, all really usefull datafiles (Raid 5)

Thanks a lot!

Rick

• Posted via NewsOne.Net: Free (anonymous) Usenet News via the Web ----- http://newsone.net/ -- Free reading and anonymous posting to 60,000+ groups NewsOne.Net prohibits users from posting spam. If this or other posts made through NewsOne.Net violate posting guidelines, email abuse_at_newsone.net