Archives of the TeradataForum
Message Posted: Wed, 20 Apr 2005 @ 08:17:00 GMT
I realize there could be some technical details of the current implementation of partitioning that are difficult to describe in a short message, and those details would make the statement clear. Having said that, I must confess I don't understand the global logic here.
Uniqueness checking has always been and still is an issue of concern with UPIs, regardless whether the table is partitioned or not. Now, I thought partitioning was meant as, among other things, a performance improving feature. The formal logic (not concerned with the technical details mentioned above) would imply that the uniqueness enforcement would rather benefit from partitioning than not. Am I wrong?
Of course, performance improving features bring some overhead with them. SI, for example, help improve performance of certain queries but require "maintanance" during DML on the table. Stats help the optimizer but their collection costs valuable production time, etc. Interestingly, partitioning seems to collide with UPIs. Good to know that. Even better would be to understand why.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|