Re: CBO vrs Rule

Hi.

 A couple of weeks ago somebody posted in that group message about  documented BUG in Oracle 7.3 CBO.
 The bug appears when analyzing less then 30% of data ( estimation  is not working so well for lesser amount of data ).  You wrote about 1% difference in stats, but did you compare  column and index stats as well?

 HTH. Michael.

In article <38bd39de$0$3481_at_wodc7nh0.news.uu.net>,   "Christopher Spence" <nospam_at_cspence@delphi-tech.com> wrote:
> Anyone have any experience with problems with CBO in 7.3.x?
>
> Full schema Analyze, 25% Estimate Statistics
> 3 days later
> 3 tables join queried with hint /*+ rule */
> Results: 240ms
> 3 tables join queried with no hints (use CBO as there is statistics)
> Results: 2300ms
> Manually Analyze the 3 tables
> 3 tables join queried with no hints (use CBO as there is statistics)
> Results: 220ms
>
> Comparing statistics between 3 days ago and today, the data (row_num,
> avg_row_leng, blocks and such) where around 1% or even less than what
 the
> analyze did last time, very few things if anything changed in this
 schema
> since then but performance was drastically effected. The main reason
 for
> the performance change is a use of a FTS on one of the tables prior
 to doing
> re-analyze
>
>

Sent via Deja.com http://www.deja.com/
Before you buy. Received on Thu Mar 02 2000 - 00:00:00 CST