Performance criteria for relational databases in different normal forms

Abstract

With the increase in client-server and other applications that are dependent on telecommunications possibilities and the use of dispersed and distributed processing, desire for access to organizational databases will increase. The increasing demands on databases make efficient storage space and access time important issues. In this article we illustrate criteria for measuring efficiency of database access as applications expressed in different normal forms. The efficiency measures are (1) actual memory space use, (2) time use, and (3) average number of operations performed. These are tested based on an optimal data search simulation of 25 different database cases for four distinct queries in both third and fourth normal form, and actual space and time use and average number of operations required by 208 subjects who performed the same queries in third and fourth normal form. We find that in almost all studied cases where there are multivalued dependencies, the fourth normal form is more efficient. These techniques, particularly the simulation and the experimental method, may be used as efficiency measures when establishing the ideal form for a new database.