Re: Guidelines for avoid Bind Variable Peeking behavior

From: Roberto Veiga <roberto_veiga_at_yahoo.com>
Date: Mon, 1 Jun 2009 08:18:59 -0700 (PDT)
Message-ID: <44177.55280.qm_at_web35605.mail.mud.yahoo.com>

We are using Oracle 10 R2.
We did a test without histogram in a QA environment and we found a lot of performance problems in many processes. We had to fallback the change rapidly. I suppose that many statatements that use histogram to perform better start to use a worse plan.  
The idea in identifying the Hash Join is because in a recent situation the plan was changed from Hash Join to Nested Loop and the performance falls down. There is nothing wrong with Hash Join! My idea is to use Hash Join and freeze this plan.  
Thanks a lot.
 
Veiga

• Em dom, 31/5/09, Wolfgang Breitling <breitliw_at_centrexcc.com> escreveu:

De: Wolfgang Breitling <breitliw_at_centrexcc.com> Assunto: Re: Guidelines for avoid Bind Variable Peeking behavior Para: roberto_veiga_at_yahoo.com
Cc: oracle-l_at_freelists.org
Data: Domingo, 31 de Maio de 2009, 11:32

What Oracle version?

I am curious to know more how the situation became worse without histograms. The plans shouldn't change as often without histograms. What is wrong with access plans using hash joins? I often prefer those over nested loops.

At 08:14 AM 5/31/2009, Roberto Veiga wrote:

>We have already tried to remove histograms and the situation become worse.
>
>-Access plan using Hash Join
>

Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com

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





      Veja quais são os assuntos do momento no Yahoo! +Buscados
http://br.maisbuscados.yahoo.com
--
http://www.freelists.org/webpage/oracle-l

Received on Mon Jun 01 2009 - 10:18:59 CDT