Archives of the TeradataForum

Message Posted: Fri, 11 Feb 2005 @ 17:14:02 GMT

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

Brian Jones wrote:

For columns with equal distribution (for skewed distribution it's hard to say):

- Distinct spools and redistibutes all rows, then eliminates duplicate rows -> efficient if the number of rows per value per AMP is low

- Group by eliminates duplicate rows in the first step locally, then redistributes the result sets and eliminates duplicate rows a second time -> efficient if the nuber of rows per value per AMP is high

- If PI-columns are aggregated ("computed locally" in explain) then Group By is always more efficient.

- And don't forget that a subquery is automatically distinct, but using Group By overrides that.

Dieter

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