A performance analysis of view materialization strategies for general expressions

Abstract

Efficient processing of views is critical to many real world applications, particularly real time applications such as surveillance systems which support military applications. This thesis compares the performance of three view materialization strategies: Semi-materialization, full materialization and query modification. This thesis first develops a program that generates databases according to user specification. Second the generated databases are used to conduct an empirical study on the three view materialization strategies using select-project-join and general expression views. The results of the study indicate that for select-project-join definitions, semi-materialization performed best for higher values of P, lower vales of 1, fv and all values of fq with the database stored on hard disk. Full materialization performed best for lower values of P, 1 and all values of fv with the database stored in RAM. The results also indicate that the semi-materialization strategy is the best view processing method for general expressions.