Archives of the TeradataForum
Message Posted: Fri, 22 Feb 2002 @ 23:12:41 GMT
My experience is that this shouldn't be a problem. At the site at which I am currently consulting, we just went through a very similar exercise and achieved significant performance improvements with little or no batch performance degradation. We achieved not only good performance improvement, but a much lowere need for spool because of the elimination of the row redistribution/spool duplication. I've been told that 4000-5000 rows per NUPI values is a good metric, although I think your rows size will affect that significantly. If the rows per NUPI is higher, you could use Multiset to still achieve good performance. However, if you are relying on the DBMS to enforce "no duplicate rows", this won't work.
Hope this helps.
Thomas F. Stanek
|Copyright 2012 - All Rights Reserved|
|Last Modified: 21 Mar 2013|