Archives of the TeradataForum
Message Posted: Thu, 12 Jul 2001 @ 11:42:05 GMT
(assuming stats are collected on all columns and NUSIs involved in the join plan and filter conditions)
Just a thought...
Seemingly, what we want is for the 2 small tables to be joined via a cross-product join then duplicated on all amps. Once this is complete,then we should be able to sort the large table (amp local) based upon the 2 Join columns and join the 2 spool files.
Try creating a 2 column NUSI on the large fact table for the 2 columns used to join to the 2 small tables. Be sure to collect statistics on this new 2 column NUSI.
If the cross-product of the 2 small tables still doesn't occur first, then you may want to try forcing this cross product in a derived table, then join the derived table to the large fact table.
Hope this helps,
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|