Re: create cluster ... nologging

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Thu, 11 Jul 2002 11:19:31 +0100
Message-ID: <1026382716.29854.0.nnrp-01.9e984b29@news.demon.co.uk>

The operations that can take advantage of NOLOGGING do all their work about the high water mark, and then bump the HWM to make the data appear.

If you have clustered tables, then the table rows have to go into the correct block in the cluster (rather like the updates to an index).

In the general case, this would make it much harder, and less efficient, to do nologging creation of blocks: the code path would have to work along the lines of:

does a row in this table have the same cluster key as a row in any of the tables already in the cluster.

If yes, then either:

put the row into the right block => logging or

        put the row into a new block, but go back
        and chain the original block to the new
        block if, and only if, the user commits.

--
Jonathan Lewis
http://www.jlcomp.demon.co.uk

Next Seminars
        UK            July / Sept
        Australia      July / August
        Malaysia        September
        USA (MI)        November
http://www.jlcomp.demon.co.uk/seminar.html

Marcin Buchwald wrote in message <3D2D57D7.DEC62D10_at_agora.pl>...


>Is it possible to create a cluster with nologging tables?

>It seems it is not.

>

>What's the reason for this?

>

>Regards,

>

>        Marcin

>

Received on Thu Jul 11 2002 - 05:19:31 CDT