Archives of the TeradataForum

Message Posted: Wed, 01 Jun 2005 @ 12:43:18 GMT


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

Subj:		Re: Why does a badly skewed join run that badly?

From:		Ole Dunweber

Hello Daniel

Thank you for answering - I've almost forgot that I posted the question...

I understand, that volatile tables can cause awfull plans (given that random Amp sampling can't exclude awfull plans).

But I still don't understand, that a badly skewed merge join runs that bad. As I wrote in the question, shouldn't a totally skewed merge join of a volatile table with 500.000 rows on 42 amps amount to a perfectly distributed merge join of a volatile table with 22 million rows?

The runtime certainly indicated, that this was not true in the particular case (the query was almost alone on the system for around 10-20 minutes).

So, maybe it's because the join is an outer join, or maybe because we run priority scheduler - which might get "confused" when a query is heavily skewed?

Anyone else having experienced something similar?

Best Regards

Ole Dunweber
Coop Nordic


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