A NOTE ON A QUOTE

Those following my writings know my position on denormalization for performance, namely that any claim to the effect that normalization – purely logical design – causes poor performance, which is determined entirely at the physical level – is based on the logical-physical confusion prevalent in the industry. Here’s some of the places I discuss this issue:

When I recently came across this week’s quote thought I should contrast an argument based on a good grasp of fundamentals and one underlied by the confusion, each leading to a different conclusion as to what the problem is, who should resolve it and how. In this note I will translate the quote, which is misleading, into the correct formulation.

Here is what Kevin Loney says in his book ORACLE 8i DBA HANDBOOK:

"...overly rigid adherence to relational table designs will lead to poor performance ... User-centered table design, rather than theory-centered table design will yield a system that better meets the users' requirements.”

And here is what he should have said, and would have, had he known and understood the relational model, and not confused levels of representation:

“Oracle, like all SQL DBMSs, maintains a one-to-one relationship between logical rows and physical records and, thus, must at all times preserve and rely on a fixed physical order of rows and columns. This limits optimization and performance may well suffer if databases are correctly normalized to minimize the risk for corruption by minimizing redundancy and the integrity burden it imposes. Users, therefore, may have to trade-off integrity for performance, which would not be necessary with a true relational implementation that adheres to the theory.”

Quite a different picture, isn’t it?

Posted 08/30/02