The problem with denormalization.
Date: Mon, 27 Sep 2004 08:11:16 -0400
Message-ID: <jKqdnbY-CITgnsXcRVn-ig_at_comcast.com>
I recently ran into a place called the "Database design Blog". In the first article I scanned, I read the following:
"Normalize until it hurts, denormalize until it works."
There is some truth to this saying, but it suffers from the same problem as all the other advice about denormalization: it doesn't tell you what to do. It reminds me of the advice on how to get out of Witt's end: "don't go west".
Designers need better advice than "denormalize". In my particular case, I've gotten a chance to design several "reporting databases" over the span of consulting for many clients. Ever since I learned star schema, I've used it wherever a reporting situation called for that type of database design.
My point here is not how wonderful star schema is. That's a different discussion.
My point is that star schema provides a positive direction to the design
process. If you reach the point where you've decided to denormalize, and
you say, "now what do I do?", you need more answers than "don't normalize".
Before I learned star schema, many of my denormalized databases were
somewhat undisciplined. Using star schema allowed me to follow an organized
system for design, instead of merely not following the trail of
normalization.
"Don't normalize" is as vague as "don't go west."