Re: estimate statistics

From: Venkat <vreddy_at_no_spamix.netcom.com>
Date: Thu, 18 Jun 1998 19:55:01 -0400
Message-ID: <6mc9c5$al4@sjx-ixn11.ix.netcom.com>

In my experience 20% is what you should use. 5% is not very representative and the optimizer might end up making terrible decisions.

Venkat

>RE: DBMS_UTILITY.ANALYZE_SCHEMA()
>Any thoughts on what is a sensible percentage
>to use for estimate statistics? Is 5% enough
>to allow the optimizer to make sensible decisions?
>OR I could use 10,000 rows, and that would be more rows
>than most of my dimension tables (I presume 10,000
>rows means 10,000 rows or *up to* 10,000 if that
>many exist) - However, 10,000 rows is only a tiny
>fraction of my fact table.
>Any thoughts, experiences?
>Steve
>
>--
>"The floggings will continue until morale improves."
>
Received on Thu Jun 18 1998 - 18:55:01 CDT