Archives of the TeradataForum

Message Posted: Fri, 05 Sep 2008 @ 18:31:11 GMT

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

Hi David,

this product join might be the least expensive plan:

For a real product join there would be no need for a redistribution of the large table, but product joining 1.632 rows to 80.000.000 requires > 130.000.000.000 comparisons.

For a Merge Join there will be the redistribution followed by a sort. This sort is quite costly.

So the current plan avoids the sort and due to the redistribution it's reducing the number of comparisons, e.g. if you got an 100 AMP system there will be ~16 rows per AMP, this results in only 1/100 of the comparisons.

just my 2 cents :-)

Dieter

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