Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: System Statistics and the CBO

Re: System Statistics and the CBO

From: Wolfgang Breitling <breitliw_at_centrexcc.com>
Date: Fri, 17 Jun 2005 09:09:29 -0600
Message-ID: <42B2E7A9.8020708@centrexcc.com>


Mladen,

That's something I was wondering myself. My statement below may very well be a bit too broad. I should rephrase it to "From all I know, the CBO reverts back to the prior IO costing model if mreadtim < sreadtim".
The cpu costing part of the new CBO cost algorithm may still be used. On the other hand I am not sure how much trust to put into what DBMS_XPLAN.DISPLAY is telling me. I have had cases where suddenly it reported CPU cost percentages without system stats being present. Finally, my experiences are mostly with 9iR2 and things like that could work slightly, or not so slightly, differently in 10g.

Mladen Gogala wrote:
> Wolfgang Breitling wrote:
>

>>

> Wolfgang, when I do SELECT * FROM TABLE(DBMS_XPLAN.DISPLAY) it tells me
> that "CPU costing is ON, despite
> the fact that I fiddled with sys.aux_stats$ and changed sreadtime to be
> mreadtime+1. Is DBMS_XPLAN lying to me? The
> version is 10.1.0.4 on Linux.
>
-- 
Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com
--
http://www.freelists.org/webpage/oracle-l
Received on Fri Jun 17 2005 - 11:14:47 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US