Archives of the TeradataForum
Message Posted: Tue, 16 Jul 2002 @ 10:10:54 GMT
Based on the Physical Implementation course, you should spend alot of time deciding the Primary Index. i.e. UPI or NUPI.
However reading the forum, I have noticed that all of the DDL that has appeared recently (e.g. SBC) always has UPI's defined.
At present we have been very selective, and have tried to envisage the kinds of questions being asked, and therefore cut down the amount of redistributions being performed, by using NUPI's on hierarchical structures.
We certainly have noticed that when using NUPI's we have alot more DBA work to take into account, to ensure that the skew of the table does not go over a certain level, and in alot of cases, our data changes over time and skewing happens. Also when the number of Nodes are increased, we have to revist all of the NUPI decisions.
What I would like to find out, is an idea of what other people have done around this area.
Do most people just define UPI's as they are alot less work, or do people spend the extra time required defining NUPI's.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|