The Normalised vs. De-normalised data debate
Recently I've been reading a lot of "experts" on sites like Stack Exchange saying that De-normalised data structures are better than Normalised data structures for performance. The simple fact is you should know the business problem and the data before making any recommendations; and someone should test the impact of their design on large chunks of test data. A system with a few hundred rows will act very different from a system with a few hundred million rows in. End of the day you need a system that can work in a business as usual environment with no maintenance just to stop it falling over.
