Re: Speeding up the redo logs

Striping redo logs can help if your redo write sizes are large say 100- 200K. If you'd like evidence just try for example an import of a large table with small striping 32/64K on the redo logs verses without.

Best
Kyle

PS THis is pretty rare. I almost always see log writes at around 10K. Since writes are piggy backed and commits on the system will be causing redo writes thus reducing the redo write size.

In article <94ep17$eau$1_at_nnrp1.deja.com>,   Tim Cuthbertson <timcuth_at_my-deja.com> wrote:
> I have implemented a very fast redo log setup, but it is not cheap. My
> average write times to online redo logs are 0.1 to 0.3 milliseconds.
> And I wholeheartedly agree that in a write intensive database, redo
 log
> time can be critical.
>
> However, I also agree with some of the other posters that you have to
> try to work on your procedures and tuning to minimize the load you are
> forcing the database to do.
>
> I have two separate disk arrays on separate SCSI controllers to be
 used
> only for the online redo logs. Log copy A goes on one array's drives,
> Log copy B on the other array. Each log file in the online rotation
> goes to a separate drive in the array, so that log archiving of the
> previous logfile does not interfere with writes to the current online
> log.
>
> Do not use RAID striping for your online redo logs - it serves no
> purpose.
>
> Finally, I have my redo logs on Veritas Quick I/O, which eliminates
> unnecessary OS buffering that Oracle is accomplishing, anyway.
>
> The whole point of all of this (except qio) is to isolate the I/O to
> the redo log file from ALL other I/O activity.
>
> Sent via Deja.com
> http://www.deja.com/
>

Sent via Deja.com
http://www.deja.com/ Received on Mon Jan 22 2001 - 12:54:44 CST