Archives of the TeradataForum

Message Posted: Wed, 07 Feb 2001 @ 09:51:02 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Eric,

You're correct.

Although there has always been the command to re-collect stats at the table level, until V2R4, this has always been processed as a series of collect stats on each column and index. Each one is done serially, so there are no time savings by using this syntax.

With V2R4 (my understanding is that it is V2R4.0.0), re-collecting stats is done by a single scan of the table so this provides significant potential time savings.

(Back to the original question) - I don't think there's much can be done to speed up the collect stats process itself. Treat it simpy as a full-table or full-index scan and use the same tricks;

- use compression on one or more columns in the table: compressing your data makes it physically smaller and hence faster to scan. Yes there is a trade-off against cpu consumption, but that's part of your decision making process

- don't do it so often: if only 1% (or even 0.1%) of your data changes in a week, I would doubt whether you need to re-collect stats every week.

- don't do it (I'm serious): do you need stats on this column or index?

Regards,

Dave

	<Prev	Next>		<<First	<Prev	Next>	Last>>