Re: intermittent "pulse" while accessing 7.3.3.1 instance

Also try increasing the number of DBwriters you have. In your init__.ora DB_WRITERS=N where N should be the sum of 1 for each 50 users AND 1 for each two batch processes. That is a minimum of two should be used. Please see page 180 of the new
"Advanced Oracle Tuning and Administration" book from Oracle Press. Many other items may be causing this -- bad use of the temp area, OS swapping etc.

David wrote:

> Shirley,
>
> This type of behaviour often occurs while the ORACLE Server is taking
> a
> 'checkpoint'.
> Make sure you have 'checkpoint_process = true' in your init.ora file.
> Also set log_checkpoint_interval to a very large value and possibly
> your
> redo logs need to be increased in size as checkpoints are taken when a
> log
> switch occurs.
>
> Also make sure 'db_files' parameter is not set to too large a value as
> this
> can reduce the efficiency of the 'db writer' process by increasing its
> scan
> depth too much.
>
> regards
> David Morgan-Brooker
>
> Shirley Barger <sbarger_at_igc.apc.org> wrote in article
> <sbarger-0208971640050001_at_news.sirius.com>...
> > We've just moved our production instance from 7.1.6 on an HP I70
 (HP-UX
> > 9.04) to 7.3.3 on an HP T520 (HP-UX 10.20). We run a widely-used
 business
> > package designed for colleges against this db, and we've run the
 same
> > version of the package for several months on the lower grade
> > machine/OS/Oracle versions. What we experience on the new platform
 are
> > intermittent pauses. Pauses last anywhere from 2 to 40 seconds, and
 they
> > appear when users are moving between forms or blocks. But when a
 pause
> > occurs, ALL Oracle users are affected--it's as if the system pulses.
 

> > Forms users and sqlplus users all see it. Unix, however, hums along
 just
> > fine--in tests, we've watched folks happily running Unix apps and
> > utilities while the Oracle folks are dead in the water.
> >
> > We have very few users right now; we are barely using our CPUs, i/o
 is
> > minimal. We were careful about tuning the kernel parameters,
 setting up
> > enough swap, and balancing the load on the RAID arrays per Oracle's
> > recommendations. We did not, however, go raw (and would prefer not
 to).
> > Our DBA can find nothing suspicious when monitoring the instance.
> >
> > Any and all suggestions on what's wrong, what to check are welcomed.
>
> TIA.
> > Shirley Barger <sbarger_at_ccsf.cc.ca.us>
> > City College of San Francisco
> >

--
eric kamradt
hm: (801) 567-0912
mailto: kamradt_at_inconnect.com
http://www.inconnect.com/~kamradt/resume.html