Archives of the TeradataForum

Message Posted: Mon, 31 Oct 2005 @ 15:22:55 GMT

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

That second formula is in Brian's book as well, however for those systems with a large number of amps, the result set can be unwieldy. Certainly a sort order helps to highlight a really lumpy (skewed) primary index.

I like the first query, because you can quickly determine the affect of a different primary index selection. The fact that you can't always achieve ratio of "1" doesn't mean the data will be "skewed". When you look at what Teradata thinks about skew, is really relative to the number of rows in the table and the number of amps. You can never achieve "perfect" distribution unless the number of actual rows in the table is evenly divisible by the number of vprocs. However, the variances of uneven distribution use an acceptable variance or standard deviation. Therefore, zero skew a "relative" term not an "absolute" one.

Michael E. McBride, MSCIS

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