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: Estimate statistics

Re: Estimate statistics

From: John Bechtel <john.bechtel_at_teldta.com>
Date: 1997/03/19
Message-ID: <333024C2.FB3@teldta.com>#1/1

> Louis Felder wrote:
> >
> > I have heard rumors that there is a bug in the esitmate statistics part
> > of analyze table in versions of Oracle prior to 7.2 Is there any truth
> > to the rumor? If it is true, does anyone know of a work around other
> > than doing a compute statistics?
> >
> > TIA
> > Lou

Marc wrote:
>
> Yes, there are several bugs with analyze even in 7.2. However, depending which
> problem you are encountering there are different work-arounds. The best method
> is to use 'compute'. However, this is also buggy for very large tables (> 6 Gb)
> so that you can't use compute at all. Then, the only work around is to estimate
> with approximate 25-30% and do manual updates to the tab$ table to have correct
> stats. If you need more information you can e-mail me directly. It is not simple
> and the best choice is probably to use RULE based optimization.
>

We have a site running on 8 cpu's, with some very large tables where they have set the degree of parallelism at the table level. Doesn't this force the cost based optimizer to be used when these tables are accessed? If so, there is no way around needing to gather statistics. Can you post/mail how you do the manual updates to the tables. The database mentioned above is 7.2.3, but we have many others from 7.0.16 -> 7.2.3 and will soon be at 7.3.

Thanks

Received on Wed Mar 19 1997 - 00:00:00 CST

Original text of this message

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