Archives of the TeradataForum

Message Posted: Fri, 07 Dec 2012 @ 14:21:39 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Hi Kumar - choice of PI depends on a couple of considerations:

1) Which PI best distributes the data? Note, modest skewing is usually a non-issue

2) How is the data accessed? To use the PI, all attributes in the PI must be qualified in the various constraints

3) How is the data loaded? Is there a tight window? Etc.

The optimal solution embraces these constraints. However, there are times when the choice of PI must weigh heavily in favor of one of the choices above at the expense of the other(s).

Rgrds,

Bill Gregg
Teradata DBA; Data Warehousing, BI Team

	<Prev	Next>		<<First	<Prev	Next>	Last>>