Re: rm RULE based optimizer != GOOD IDEA

From: Gints Plivna <gints.plivna_at_gmail.com>
Date: Mon, 25 Apr 2005 11:41:24 +0300
Message-ID: <6e49b6d005042501417df9a459@mail.gmail.com>

One of our developers used RULE hint widely in our latest project. As a result we had to inspect almost all of them (taking time and $$$) when moving to production. So from now I'm planning to allow RULE then and only then if developer can explain _ALL_ RBO rules :))) (http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96533/rbo.ht= m#38864,
http://www.oreilly.de/catalog/orsqltunpr/chapter/)

As a result I hope that those who will be able to do that would understand that it is a bad idea.

Gints

On 4/22/05, Robyn <robyn.sands_at_gmail.com> wrote:
> I for one will be happy to see the rule based optimizer go. Most of
> our developers and dbas add rule hints as a quick fix without taking
> the time to actually figure out the problem. Then, they announce
> that the query completed 'really fast' in dev and move it to
> production, where it may or may not run 'really fast'. Either way,
> there is no baseline to work from if we need to continue to tune it.
> Plus, the rule hint is likely to exist through the next 3 upgrades,
> and no one remembers that it exists, much less knows if it's still
> making things 'really fast'.

>=20

> As a result, we have lots of legacy queries that need to be
> reevaluated and there have been many times when REMOVING the rule hint
> in an old query sped things up considerably.

--
http://www.freelists.org/webpage/oracle-l

Received on Mon Apr 25 2005 - 04:45:38 CDT