Re: Performance Diagnosis

"Michel Cadot" <micadot{at}altern{dot}org> a écrit dans le message de news:40d72d12$0$15195$636a15ce_at_news.free.fr...
>
> "Matt" <mccmx_at_hotmail.com> a écrit dans le message de
> news:cfee5bcf.0406210131.46b97020_at_posting.google.com...
> > > From Steve Adams's ixora site:
> > >
> > > rdbms ipc reply waits indicate the either DBWn's write batch is too
> > > big, or you are checkpointing too intensively. Make sure that db_files
> > > is not much higher than the number of datafiles that you actually
> > > have. If you are using incremental checkpoints (
> > > db_block_max_dirty_target) don't be too ambitious.
> >
> > As far as I was aware, the db_files parameter had no effect on the
> > DBWR write batch size in 8i (can't access ixora at the moment to
> > confirm).
>
> _db_block_write_batch defaults to the least of:
> . db_files * db_file_simultaneous_writes / 2
> . db_block_buffers / 4
>
> --
> Regards
> Michel Cadot
>
> >
> > I haven't set db_block_max_dirty_target explicitly so this is the
> > default value.
> >
> > Matt
>
>

Forgot to mention: don't decrease this parameter to a too low value or you'll get "free buffer wait" events.
I think the best way is to set db_files to a value slighty higher than the number of files and adjust db_file_simultaneous_writes to a value that set _db_block_write_batch to a value higher than the current DBWR batch size ("physical writes"/"write requests" in v$sysstat).

-- 
Regards
Michel Cadot