Re: Oracle 7.0NT-Optimizer, full of bugs?

In <jdennisCwox72.9Io_at_netcom.com> jdennis_at_netcom.com (John Dennis) writes:

>
>
>Quick note on this subject:
>
>If a index is not being use when it should, delete the statitics on
>*just* the index in question. Oracle optimizer should then pick it
>up and run with it. I have no idea why this works.
>
>BTW, I greatly dislike (hate) the comment - "just use rule based".
>Like if your having a problem with some bug in stored procedures
>and someone says "just use an external C program instead". Not
>necessarily good advice. If you have truly dynamic queries then
>the optimizer is your best bet with some manipulation (deleting) of
>some analyzed info and hints on just those really bad queries and
>upgrade to 7.1
>
>Yea 7.0 optimizer is a rough road but don't turn around - paved road is
>just ahead (7.1).
>
>John Dennis
>
>

John, I would submit that the concept of cost based optimization is a flawed concept to begin with. The overhead required to re-calculate the best path to the data as the schema and population changes greatly reduces any potential benefit. I have never seen cost based optimization work correctly. I always use rule based as a baseline, and my own intuition as the final say.

-Richard Armstrong Received on Mon Sep 26 1994 - 03:23:06 CET