Database normalization - Creating high performance from the start.

However, due to the language used in formal definitions of the different Normal Forms, many people have had difficulties in understanding the full concept of database normalization.

And I have to agree; understanding relational algebra, or the formal descriptions of the Normal Forms (which actually are neccessary to cover all aspects), doesn't make it easy.

However, the fun part of it is, that even without the theoretical conciousness of normalization theory, most people normalize their ER models intuitively: You know that the name of your customers should only appear once, and in one place; the CUSTOMERS table. You also know that for each order line, you do not have to calculate the value of the sale. All this, that you actually know, and intuitively practice, is database normalization. We humans tend to be effective, even if we are not always aware of it.

The only problem with this approach, is that when you are not operating in a methodical manner, you might slip a little here and there. That's when problems enter into your analysis and/or design: It is no longer consistent throughout the whole model.

Actually, if you participate in discussion forums, and read all messages for help on various problems, you can see that many of the problems are relating to results of poorly designed and/or denormalized databases.

In today's database theory, we operate with 5 Normal Forms; 1NF through 5NF. In addition, we sometimes refer to Boyce/Codd Normal Form (BCNF), which is a "higher grade" of 3NF, and Domain/Key Normal Form (DKNF), which focuses on database-side constraint definitions. Actually some people are now also talking about a 6th Normal Form, 6NF: Basicly, tables are decomposed into even smaller tables, containing an identifier plus one data column. I won't be discussing that further for now; maybe later ;-)

• Formal definition of the 5 Normal Forms In this keynote, I present to you the formal definitions of the 5 common Normal Forms. Quite hard to grasp? In my eBook I have used a much more understandable language and E-R examples, or at least, so I think.
• Normalization, Denormalization and Performance Denormalization is almost as popular a term as normalization. It is widely assumed that to denormalize a database is a means of computing (derived) values and store them in the database, or store the same information in several places in order to 'help' the system to perform better. This keynote makes you aware of the perhaps not so obvious dangers in such a process.

Return to Database Design home