Archives of the TeradataForum

Message Posted: Mon, 30 Jan 2006 @ 09:50:53 GMT


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

Subj:		Re: Erwin is not mi amigo again...

From:		Dunweber, Ole

Anomy wrote a couple of things, which I must comment on...

PI's are great for data distribution but make lousy unique identifiers; However, the bigger the table the better it is to have a UNIQUE PI.

Sounds contradictory but the savings can be realized by having queries utilize the PI; after all, table scans on big tables, even partitioned ones can be very costly. Best to educate folks on using a PI on the big ones all the time.

Hmm, do you really compromise the most important property on a table (the PI), for the sake of uniqueness?

How many business questions can you answer (using the PI), if you have put say 10 columns into it, to make it unique?

Can you "educate" folks (the business people I guess you mean), on which business questions they are allowed to ask?

Yes, the table scans can be costly, but star-joins are great. Sadly you wont get them, when you put a lot of columns into the PI.

NO USIS or MULTISETS allowed in my Datawarehouse.

Do you really consider the Data warehouse to be yours?

In my humble opinion, it belongs to the business.

Regards,

Ole Dunweber
Coop Nordic


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

Attachments

Library

Quick Reference