Archives of the TeradataForum

Message Posted: Tue, 16 Sep 2003 @ 18:13:28 GMT

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

Michael,

We're having some issues of late with Partitioned Hash Joins, similar to your scenario. I bet if you look at the skew by amp, you'll notice that only one or two of your AMPs are doing all the work.

In the Documentation, it notes that hash joins can result in degraded performance if the data is badly skewed. If so, you can turn them off or change the skew allowance to 80 or 90.

Best bet, if this is a skewing issue, determine what your two spool files (9029 and 9030) have been hashed on for the join in step 41. If possible, collect stats on the repsective columns. Hopefully, with the proper stats, the optimizer will realize that this join decision is a poor choice.

God Luck,

Clay

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