RE: Was 10G and UFS - long write times (is, I/O really?)

From: Kevin Closson <kevinc_at_polyserve.com>
Date: Thu, 8 Jun 2006 13:45:49 -0700
Message-ID: <5D2570CAFC98974F9B6A759D1C74BAD0E5A733@ex2.ms.polyserve.com>

>>>

>>>Kevin, I am not sure what your email is answering (I am
>>>missing the quoted question) but wouldn't it be better in
>>>10g to test with "commit batch nowait" rather then turning
>>>of logging with the underscore parameter ?

... if you want to totally rule out the cost of LGWR physical I/O you have to set _disable_logging=TRUE. This is most concise way of putting blame on the actual LGWR I/O component of some foreground log file sync. All the other code involved with a LFS is still executed.

Don't do this with a database you really care about, and only do shutdown normal of course.

--
http://www.freelists.org/webpage/oracle-l

Received on Thu Jun 08 2006 - 15:45:49 CDT