Archives of the TeradataForum

Message Posted: Tue, 31 Jan 2006 @ 15:17:44 GMT

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

<-- Anonymously Posted: Tuesday, January 31, 2006 10:16 -->

Victor,

Maybe it is because we are talking about different things. In my view, there's a difference between LANDING and STAGE tables and DATAWEARHOUESE tables.

The data in LANDING and STAGE can be in any form. Datatype differences, denormalizaed information, duplicates, etc. The use of a MULTISET table may be OK in that instance to expedite the process. The table definitions in LANDING and STAGE have to be flexible in order to reflect the source world. Most of the tables in those two databases are nothing but flat files loaded into tables, hardly a relational concept is it not?

The actual DATAWAREHOUSE database is a completely different issue. MULTISETS do not have a reason to exist in the DW because they are not relational and they will yield wrong results if duplicates do exist. Yielding the wrong results or attempting to find a DUP on a 20B row table are two risks I am not willing to accept in my DW. I hope we can disagree on that.

The ultimate responsibility of the DW well being rests with the Data Administrator, that being me. I have been around the block long enough to have seen my fair share of "IT blunders" occur. Not allowing MULTISETS in the DW is one way of reducing the DW exposure to said blunders.

I think enough has been said on this topic that many are tired of it. I hope we can agree to disagree on the value and utility of MULTISETS and I hope we can learn to respect that.

Anonym

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