Re: Rule vs. Cost

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Thu, 11 Nov 1999 09:05:25 -0000
Message-ID: <942311206.14373.1.nnrp-07.9e984b29@news.demon.co.uk>

This is a side-effect of the cost-based optimiser in CHOOSE mode favouring tablescans and hash-joins where the RULE-based might be required to use an index.

Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk

attwoody wrote in message <80d9dd$tbo$1_at_nnrp1.deja.com>...
>
> The other thing I noticed was that when the OPTIMIZER_MODE was set
>to RULE, the individual buffer hit ratios were about 99-100%, overall
>about 98%. CONSISTENT_GETS were in the hundreds of millions. When I
>changed it back to CHOOSE, individual buffer hit ratios went down a
>bit, with most being between 97-100%, but the overall hit ratio went
>down to about 80% (it had never gone below 90% when the mode was
>RULE). CONSISTENT_GETS went down to the hundreds of thousands.
>
Received on Thu Nov 11 1999 - 03:05:25 CST