 |
 |
Archives of the TeradataForum
Message Posted: Tue, 29 Apr 2003 @ 16:51:31 GMT
| Subj: | | Re: Collect statistics.. |
| |
| From: | | John Hall |
One thing about dropping statistics when they are found to be a problem: Contact NCR and have them treat it as a bug which should be
addressed. The problem with the expediency of simply dropping the statistics is that dropping them for one class of queries leaves an
exposure for other queries. If it's not addressed, you'll constantly be trying to tune queries that the parser/optimizer should have gotten
right in the first place.
With the feedback from the customer (which means that you'll have to document and pursue the problem), NCR should be able to make further
improvements in the parser/optimizer. I know that I'm being somewhat contrary, but I would rather have a rock-solid product than new
features and if the optimizer can benefit from statistics in most situations, then it should benefit from them in all situations - they
should never be detrimental.
| |
