Home Page for the TeradataForum

Archives of the TeradataForum

Message Posted: Fri, 08 Nov 2002 @ 22:04:49 GMT

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

Subj:   Re: 3NF vs Dimensional modelling
From:   McBride, Michael

Well, I did not want to go down this path, but I feel forced into in, if only because I am a hard-core, diehard, 3NF implementation bigot from way back. I have read all the Kimbal idealogy that had, at its core, the one ulterior motive, to sell his own endorsed DBMS products.

First off, I have never been in favor of ODS, as a part of a EDW, but I am in favor of Active Data Warehousing, and the data WILL always get into the DW faster in 3NF than in the DM, that my friend, is a fact.

Point of fact, well designed 3NF models based on the true business subject matter will perform at the expected, reasonable performance criteria 90-95 % of the time.

The key principle here is that the implemented model 'correctly' reflects the business (BQ, and BR) in all areas. I have seen horrible business models that have perfect 3NF implementation as applied to technique and theory. But the model did not reflect the business, this is almost always the case when I here statements like the ones you made such as: "One very major problem with the 3NF is that it isn't very useful for satisfying queries directly" and "Because no one can query the EDW directly, or at least, not easily or efficiently" and "So I would judge the 3NF approach on what it can and can't support architecturally. In the case of intra-day reporting, it's dead in the water."

How open minded is that last statement? I would cetainly require verification of that one, since I have replaced many DM with 3NF implementations that out-performed the DM for all but the most basest of "canned" queries (for which the DM was tuned for). And I have improved data availability by hours in every case.

Bottom, line, the bad-rap that 3NF has gotten in the DW industry, is because too many people trained in a technique do not understand how to apply the technique to the art of business modeling. One can learn the techniques to paint oils on convas, but it takes talent and experience to be considered a master artist. (feel free to quote me on that one, I made it up on my very own a long time ago) Basically, the fault usually lies in the model itself and not the technique used. Granted, it is easier to mask a missing business rule in a huge fact table with tons of dimensions, but, as harsh as it sounds, I consider that just plain lazyiness.

So, more to the point, the EDW is only as good as the foundational model (or modeler), both in technique AND functionality, as the case may be.

I will say that DM has a place and purpose (niche) in a Total Data Warehouse Solution, just not for storing the Corporate Data Assets at the core of an Enterprise Data Warehouse.

(I'm quite sure I'm closed minded on that thought though)

...and the 'religious' wars go on... and on... and on...

...but ya' gotta love it!


  <Prev Next>   <<First <Prev Next> Last>>  
  Top Home Privacy Feedback  
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 28 Jun 2020