Re: Oracle Cost Based Optimizer

From: Darby Crash <nospam_at_satan.com>
Date: 1997/07/28
Message-ID: <01bc9bbb$808ac920$5cf32399_at_emailid-pc.Cisco.com>#1/1


Cost-based is much better in 7.3. Since you can "turn it on" globally via the init.ora, it might be worth a try for you to check it out. The basis
of the cost-based optimizer is that when you run ANALYZE ( I use "exec dbms_utility.analyze_schema(<SCHEMANAME>,...)" to analyze the entire schema at once), you generate a bunch of statistics that the optimizer can use to determine the optimal execution path for the query. Since the optimizer knows how many rows are in the tables you are joining, it can decide which indexes to use, etc. I would recommend the book "Oracle Performance Tuning" on O'Reilly pub. They just came out with this huge new issue - very informative and pretty cheap.

jeremy_at_cs.unm.edu wrote in article <869155958.18595_at_dejanews.com>...
> I have a certain question, and currently I have not been able to find the
> answer either way on this one.
>
> I'm working on a project that involves Oracle v7.3.2, right now we are
> using rule based optimization. I have heard horror stories about Oracle's
> CBO, (i.e. not picking the optimal path for a query, not returning.. a
> very undesirable feature). And most of the consensus of other parties
> is that if you use a version of Oracle before v7.3 then USE RBO OR
 ELSE!!!
>
> I have heard rumors that the CBO will be fixed in v7.3 and above, and I
> have heard rumors that the CBO won't be fixed until v8...
>
> Does anyone know which is it? or is anyone running v7.3 and the CBO with
> no problems?
>
> Also... I have read the Oracle documentation on their Optimizer... Does
> anyone know where I can get a hold of some material that covers the
> analysis of the Oracle optimizer... maybe as compared to other optimizers
> (i.e. INGRES)...
>
>
> -Jeremy
>
> -------------------==== Posted via Deja News ====-----------------------
> http://www.dejanews.com/ Search, Read, Post to Usenet
>
  Received on Mon Jul 28 1997 - 00:00:00 CEST

Original text of this message