Archives of the TeradataForum

Message Posted: Tue, 02 Jul 2002 @ 20:03:41 GMT

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

This matches some of the problems we've seen at Boeing as well.

In simple form, you have a large table, a small table and a set of where clause conditions that are highly selective. The old plan was retrieve from large table using where conditions producing small subset, redistribute subset and small table for join. The new plan is redistribute both tables and do join, applying where conditions as join conditions. This can result in much larger spool files. (BTW, this scenario is usually embedded in a much larger query).

Workarounds include putting the large table and related conditions in derived table, extracting partial result sets to (global) temporary tables, and/or adding statistics & indexes on columns used in where clause. No two cases are exactly alike, so it's time consuming fiddling with the workarounds to find one that works. And there's always the worry that an exotic workaround will penalize you on the NEXT upgrade.

/david 'tweak that query' hough

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