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

Home -> Community -> Usenet -> c.d.o.server -> Re: Query CBO Cost - Hints

Re: Query CBO Cost - Hints

From: Niall Litchfield <niall.litchfield_at_dial.pipex.com>
Date: Sat, 9 Apr 2005 14:59:28 +0100
Message-ID: <4257df7c$0$293$cc9e4d1f@news-text.dial.pipex.com>


<bdbafh_at_gmail.com> wrote in message
news:1112995246.090664.288030_at_f14g2000cwb.googlegroups.com...
> Costs across parsings don't necessarily mean anything.

Hi Paul

They *might* be incomparable, but normally they are comparable within the same system. I've recently seen a number of posts - and a couple of metalink documents - that backup Jonathan's assertion that the cost is a direct estimate of the elapsed time the query will take (in units of the average single block disk access time). Given either no system stats, or consistent system stats - neither a big leap for 9i, then I cannot see any reason why you cannot compare costs over parsings, in the same system. The same plan may have a different cost because of differences in the stats, but the costs *should* in my view be comparable - the only thing that I can think of that should make a difference is a change in system stats.

If anyone has done any work comparing actual elapsed times to cbo costs, I'd be extremely interested to see it. If not I guess there is some work to be done.

-- 
Niall Litchfield
Oracle DBA
http://www.niall.litchfield.dial.pipex.com

 
Received on Sat Apr 09 2005 - 08:59:28 CDT

Original text of this message

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