|
|
Archives of the TeradataForum
Message Posted: Fri, 29 Jan 2010 @ 16:20:00 GMT
Subj: | | Re: Workaround to avoid product join |
|
From: | | Dieter Noeth |
Victor Sokovin wrote:
| Product join has the best performance? I guess we all are learning something new here every day. | |
I wondered, too.
It's not the product join per se, but the huge intermediate spool created by it. An average range of just 10 minutes already results in
1,500,000,000 rows (150,000,000 * 10).
I compared both queries on about 10 mio random rows with 3 products and approx. 12 minutes average range and there was a 1:40 difference in
runtime and CPU usage.
| A lot depends on the data, of course. Perhaps the pre-aggregation suggested by Dieter did not reduce the volumes much in your
particular case? The skew is another thing to check, as always. | |
Yep, would be nice to get some more info about the actual data.
Dieter
| |