Re: To De-Normalize or Normalize ?

Christian,

Denormalization should be one of the last arguments to get faster response times from queries.

Why?

Because a clean data model is very important for building your applications, maintaining the database's integrity and for future updates.

There are a lot of other possibilities you can check before denormalize the model. Such as:

1. Correct datamodel
2. Correct indexes
3. Tuned SQL-statements
4. Index on another tablespace than your data.
5. Optimal configuration of drives, tablespaces and tables
6. Correct storage parameters for your tables (fragmentation!)
7. Enough system memory
8. A SQA that is large enough
9. etc, etc, etc

>Why and/or When do YOU either De-Normalize or Normalize your database ?

First of all, always start with a normalized datamodel. Only denormalize if all other possibilities fail. And denormalization is not the only possibility. There's also duplication, vertical and horizontal fragmentation, clustering and bitmapped indexes for example.

Good luck!

• Martin Jesterhoudt
• martinj(replace this by @)xs4all.nl

  ---