Re: Optimizer Stats collection for Datawarehouses

From: Greg Rahn <greg_at_structureddata.org>
Date: Thu, 19 Nov 2009 09:04:17 -0800
Message-ID: <a9c093440911190904n50d788c2t25e4d824d7a8086c_at_mail.gmail.com>

If the partition is not bulk loaded in a single shot, then a good way to deal with this is to clone partition stats ( dbms_stats.copy_table_stats in >=10.2.0.4 ) from another representative partition.
See:
http://optimizermagic.blogspot.com/2009/02/maintaining-statistics-on-large.html http://download-west.oracle.com/docs/cd/B28359_01/appdev.111/b28419/d_stats.htm#BABDDBGH

On Thu, Nov 19, 2009 at 7:11 AM, Mark W. Farnham <mwf_at_rsiz.com> wrote:
> One more thing, my earlier response was more about adding a new "empty"
> partition to be filled up in an OLTP or batch transaction context, and in
> the context of the thread being about datawarehouses, you should follow the
> "scaling to infinity" approach of swapping the partition in after you have
> already filled it. See Tim Gorman's paper on the topic and you'll likely
> make all the right choices.

-- 
Regards,
Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l

Received on Thu Nov 19 2009 - 11:04:17 CST