Re: estimate_percent size on gather_stats

From: Sybrand Bakker <postbus_at_sybrandb.demon.nl>
Date: 23 Jul 2002 00:40:19 -0700
Message-ID: <a20d28ee.0207222340.2b9e3152_at_posting.google.com>


Shankar <shan0075_at_yahoo.com> wrote in message news:<3D38D583.7020402_at_yahoo.com>...
> Hi Everybody,
>
> What is the appropriate estimate_percent on gathering stats for huge
> warehouse kind of tables ? Compute is not a option, as it takes longer
> time to complete. Is there any math behind the total bytes used by the
> segment that can closely match and lead to optimal execution plans.
>
> thanks,
> Shan

I've read a recommendation once in Metalink that 20 percent should be sufficient.
In the past I have had the experience though estimate results are highly unreliable. dbms_stats can work with multiple processors. I have found this really helps.

Hth

Sybrand Bakker
Senior Oracle DBA Received on Tue Jul 23 2002 - 09:40:19 CEST

Original text of this message