Re: Filesystem blocksize for archive logs on AIX
From: Rich Jesse <rjoralist3_at_society.servebeer.com>
Date: Mon, 28 Oct 2013 10:38:23 -0500 (CDT)
Message-ID: <1c0d399a8ffb50e70391ae27fe63cde5.squirrel_at_society.servebeer.com>
Ls Cheng replies:
Date: Mon, 28 Oct 2013 10:38:23 -0500 (CDT)
Message-ID: <1c0d399a8ffb50e70391ae27fe63cde5.squirrel_at_society.servebeer.com>
Ls Cheng replies:
> With redo log it does make a hell lot of difference, a few months ago with
> the objective of reducinjg log file sync waits, when redo are moved to 512
> bytes FS from 4K FS log file sync time was improved at 55%, more or less
> from average 13 ms to 6 ms, but the big difference is after reconfiguring 2
> log file member to 1, 6ms was reduced to 1.5ms roughly.
So you only have 1 member per redo log group? Even with a decent SAN to provide hardware mirroring, that setup would make me nervous, having experienced file system corruption before. But the numbers are interesting.
Anyone else only running 1 log member per group in production?
Thanks!
Rich
-- http://www.freelists.org/webpage/oracle-lReceived on Mon Oct 28 2013 - 16:38:23 CET