Archives of the TeradataForum

Message Posted: Fri, 19 Aug 2005 @ 08:10:19 GMT

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

One thing I tend not to do with Dimensions is NOT have Start and End dates. I assume you have this for the 'Slowly Changing Dimension' situation, whereby you may need to report on data from the Fact table by what it's Dimensional appeared at the time of the Fact table entry.

What I used to do was to have the Dimensions as 'current picture' i.e. only one record per key. Then, only for those attributes that absolutely need to be tracked over time would I create another specific Dimension, each with a separate Integer Surrogate key. This key would then ALSO be added to the Fact table. In this way you can see both Current and Historical pictures of the data.

This was a technique I read from the Ralph Kimball school of Dimensional modelling, by the way.

Although I haven't tried it, this approach would then allow you to construct a multi-column Primary Index and allow Teradata to do a Small Table - Big Table join i.e. it prejoins ALL the Dimensions and then accesses the Fact table via the Primary Index.

If you're looking for a completely new approach, then that may be the one to consider.

In terms of Soft RI, I do have some very recent insider knowledge from the very top Teradata technical consultants (because I was discussing it with them the other day) :

It works best with Non-Ansi joins which, according to the consultant, was due to some 'architectural issues'. Somewhat restrictive this, because it's a little difficult in doing Left Joins with non-Ansi syntax, of course.

_______________________
Dave Clough
Database Designer
Express ICS

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