Archives of the TeradataForum

Message Posted: Mon, 15 Dec 2003 @ 18:39:07 GMT

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

Jon,

NCR have updated the documentation index over the weekend (great work, BTW!) and I was able to access some documents which I don't have in my local library. I'd like to go back to your statement that statistics should not be influenced by the index subtable. To these eyes this statement contradicts with the following paragraph on sampled statistics from "SQL Reference - Volume 2 Statement and Transaction Processing", 2-34, p.80 (Dec 2002):

"If the columns are not indexed, then the rows are organized randomly, so Teradata just scans the first n percent it finds, where the value of n is determined by the relative presence or absence of skew in the data.

If the columns are indexed, then more sophisticated sampling is performed to take advantage of the hash-sequenced row ordering."

I read this as though the index subtable might actually be used by this 'sophisticated sampling'.

As I said, I see a bit of a contradiction here so it would be good to clarify this point.

Regards,

Victor

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