Planning the Physical Structure of Databases

A great deal of work is done in designing a database that is only briefly described here. (Database design is a book in itself.) Someone must define the tables needed to serve the business function that the database must support. Usually this is done through a process that has a logical design phase in which the entities and attributes of interest to the organization are identified, together with the relationships among entities. In this process, an activity called normalization is used to remove redundant data. Referential integrity and other business rules are identified at this time. Quite often, physical design is performed at this time as well so that high performance can be ensured. During this process, the normalized design can be denormalized to improve performance. Appendix A, "Bibliography," lists many excellent books on these topics.

This discussion about the database creation process starts at the point where all the tables and indexes in the database have been identified and the estimated number of rows in each table is known. You can create a database without this information, of course; but with this information, you can estimate what the size of the database should be. In addition, you can use this information to determine what file groups you need. The following sections explain why this is important.