Re: Cost Vs. Rules Based Optimizers

From: Jeff Perry <jp_at_vllyoak.resun.com>
Date: Mon, 16 Jan 95 02:12:12 PST
Message-ID: <2iN1yc3w165w_at_vllyoak.resun.com>


vv_at_mbunix (Vincent A Ventrone) writes:

> >
> >Has anyone done any extensive testing between the two optimizers where you
> >have switched from running highly optimized and very large queries against a
> >optimizer and most of them go against two entire years of sales history.
> > (Stuff deleted...)
> >We know there are minor differences in syntax, such as the driving table
>
> We haven't run tests yet, but according to the Oracle Tuning App.
> under v 7 class (which I took last week -- and recommend) syntax
> is supposed to have no effect on the execution plan generated by
> the cost-based optimizer (that is, order of FROM tables and WHERE
> clauses). Supposedly it uses only infomation generated by ANALYZE.
>
> Of course, the claims may be wrong....:-)

We found that the cost-based optimizer caused incorrect results. Our applications people reported no problems after switching to the rules-based optimizer (7.0.13.1).

--
Jeff Perry
jp%vllyoak.resun.com%blkhole_at_sdd.hp.com
Received on Mon Jan 16 1995 - 11:12:12 CET

Original text of this message