Archives of the TeradataForum

Message Posted: Tue, 16 Jul 2002 @ 11:52:30 GMT

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

Matthew,

I agree with Claybourne. A typical first-cut of PI choice is a common NUPI across the largest tables (more on this later) and UPI's on your smaller tables.

To take the large table NUPI discussion a bit further... What you often find in a data warehouse is that there are only a few tables (anything between 4 and maybe 15-20) which are responsible for the vast majority of the data volume on the warehouse. These few large tables will fall into a number of 'groups', maybe only 1, probably no more than 3 or 4. They are grouped based on the type of data they contain or perhaps the fact that all tables within the group will typically be join together. Making sure that the tables within each group have a common PI will often speed the joins of these tables.

As a brief example, a warehouse for a retailer may have the following groups of large tables:

-- a transaction detail and a transaction header table (which will probably be two of the largest tables on the warehouse) and these could have a common NUPI,

-- customer information tables (address, loyalty club membership, family, demographics)

-- campaign contact tables

-- etc

Just some thoughts.

Cheers,

Dave

Ward Analytics Ltd: Information in motion (www.ward-analytics.com)

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