Archives of the TeradataForum

Message Posted: Thu, 29 Jan 2004 @ 23:43:36 GMT


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

Subj:		Re: Statistics and explain plans

From:		Victor Sokovin

Jon,

I was using the word distribution in the narrow sense of how rows are distributed across the amps. You're right about that. And statistics (in the narrow meaning of Teradata statistics) contain no information about that.

Taradata statistics are global.

There is a popular misconception that you need to re-collect statistics if you reconfig to add (or remove) amps. Not true. A reconfig does not change your data, it just moves it around If your data has not changed, you don't need to re-collect stats.

Agreed on all these points and thanks for the confirmation.

I just scanned the thread again and I think your comments would have been better understood if you had posted them as a reply to the posting by Anomy on Jan 22.

The actual sequence of postings was different:

Anomy Jan 22 - James - Dieter - Jon.

I also still think that James's statement "Statistics do not recognize data skewing" is not correct regardless which meaning of the word "skew" he used. Why? Let's see: if it were in the "TD narrow sense" = "PI distribution skew", then it is _not always_ true as you, Jon, have explained yourself. Not always true means not true when one has to choose between "true" or "false", I suppose. If James used "skew" in the general mathematical sense, then his statement would never be true, period.

Dieter's remark, on the other hand, was true in the general sense and I cannot see anything in it that would suggest that he had any popular misconceptions that you mention. Hence my posting.

It would be interesting if the Anomy who published his/her comments on Jan 22 could actually provide more information about his/her query so that we can see why he/she was expecting the stats to be connected with the data distribution across AMPs. That would give more body to the thread.

Once again, thank you for your comments.

Victor


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

Attachments

Library

Quick Reference