Archives of the TeradataForum

Message Posted: Thu, 03 Mar 2005 @ 17:59:21 GMT

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

Normalized data structures are essential when promoting an Enterprise Wide Information Factory (Data Warehouse) per the Inmon model. The reason being that the most important aspect of an EDW is flexibility and consistency of information when supporting multiple business subject matter (e.g. Finance, HR, Sales, Marketing, Manufacturing, etc.) It is a proven fact that 3rd normal form data architecture provides the most stable, most reliable representation of a corporations data assets. It is also the easiest to maintain over long periods of time (history). And, considering how much information can be produced today in even a small business, also conserves the greatest amount of space. The primary goal of an EDW is to provide a '...single version of the truth...' without regard to application or process.

Sure, I agree with the basic premise that you might want to have summarized, pre-aggregated, or even redundant (de-normalized) data structures. These should be the exception and not the norm (IMHO). I typically would maintain the normalized architecture at the core of the EDW, and provide de-normalized structures as "dependant data-marts", sourced directly (and only) from the normalized, or "core", Enterprise Wide data structures.

Michael E. McBride
Teradata Certified Master
Teradata Database Administrator
Data Architect and Data Warehouse Practitioner
American Eagle Outfitters

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