RE: ORACLE automatic update statistics better??
- From: "Powell, Mark" <mark.powell2@xxxxxx>
- To: "oracle-l@xxxxxxxxxxxxx" <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 13 Aug 2010 18:26:59 +0000
How about make "always use the defaults" into "always start with the defaults.
Use what works for your environment."
________________________________
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On
Behalf Of Niall Litchfield
Sent: Friday, August 13, 2010 1:26 PM
To: greg@xxxxxxxxxxxxxxxxxx
Cc: mccdba1@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: Re: ORACLE automatic update statistics better??
Hi Greg
Very nice demo. I think the original consultant's recommendation as described
to us was to run with the default gather stats job, not the
dbms_stats.gather_xxx_stats default. IIIRC they are almost but not quite the
same. In particular the auto maintainance job tends,ISTM, to create too many
histograms. I'm not sure why - but publishing what the job precisely does and
behavior changes across releases would help *hint* *hint* :-)
Incidentally the consultant's recommendations are fairly close to what you'd
get if you hired me, namely "start with the built-in in job unless your
software precludes it (eg Apps) and deal with execution plans and object stats
on an exception basis, don't try and second guess Oracle until you need to"
which advice of course might end up summarized on a mailing list as "always use
the defaults" :)
On 13 Aug 2010 17:56, "Greg Rahn"
<greg@xxxxxxxxxxxxxxxxxx<mailto:greg@xxxxxxxxxxxxxxxxxx>> wrote:
I would strongly discourage the use of
method_opt=> 'FOR ALL INDEXED COLUMNS SIZE 1'
unless you really know what you are doing and why you are doing it.
More here:
http://structureddata.org/2008/10/14/dbms_stats-method_opt-and-for-all-indexed-columns/
On Fri, Aug 13, 2010 at 8:08 AM, dba1 mcc
<mccdba1@xxxxxxxxx<mailto:mccdba1@xxxxxxxxx>> wrote:
>
> we have ORACLE 10GR2 on LI...
--
Regards,
Greg Rahn
http://structureddata.org
--
//www.freelists.org/webpage/oracle-l
Other related posts:
