Re: DBWR Performance.

There does seem to be a difference in 8.1 with writes to temporary tablespaces (as in create temporary tablespace...) I keep finding that the number of writes is less than I expect and haven't found a good explanation yet.

Correct - if redo small copies is large (in fact more than a few percent of redo entries) than you want to reduce log_small_entry_size.

When LGWR writes out the log buffer each write has to be padded to the nearest complete (redo) block. This is the redo wastage. The value is usually 'redo writes x half a block'. NB The redo block size is not (necessarily) the same as the Oracle block size, nor the underlying o/s block size.

--

Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk

Bold wrote in message <388E0C27.1E83DC7F_at_yahoo.com>...
>Yep, my mistake on the DBWR Timeouts.....thanks
>In terms of the Temporary tablespace blocks written, I read incorrect
>information in a book then... :-0 .. imagine that.
>I will alter the "temporary" tablespace, and progressively implement a few
of
>your other suggestions (if things warrent) and see if things improve.
>If redo_small_copies are large, then reducing the log_small_entry_size
should
>help?
>
>Also, anyone know what the heck redo wastage is? I've never been able to
find
>out.
>
Received on Tue Jan 25 2000 - 15:31:22 CST