Archives of the TeradataForum
Message Posted: Sun, 05 Nov 2001 @ 02:44:30 GMT
Just my two cents to this thread. As pointed out, NCR does not recommend 3NF because it works "better" with 3NF over other modeling techniques. Teradata is more performant with any form of relational model: 3nf, dimensional, star, snow-flake, etc.
NCR recommends 3NF for the base enterprise layer because it can truly reflect the data and the business relationships between the data. With all the data present and modeled correctly, we can answer any question we want, now, without having to change or add to our dimensional model which is designed to answer specific questions.
From the 3NF model, we can provide alternate views, virtually or physically as needed for any toolset. This combination gives us the ultimate flexibility, for the lowest cost over time (as we don't need to redo ETLs to obtain new data not included in the dimensional model).
And, best of all, this works :) I have implemented virtual dimensional overlays on mulitple warehouses now for Micro Strategy reporting.
David W. Roth
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|