Analysis of storage structure performance for multiple data processing activities

Abstract

The File Design Analyzer, an analytical model for evaluating storage structures, has been applied to the problem of comparative performance with respect to individual data processing activities ranging from random retrievals to batch update and complex query specification. Performance is defined in terms of I/O processing time so that meaningful quantitative tradeoffs for multiple data processing applications can be determined. The model has been extended to further evaluate storage structures in terms of three levels of secondary storage contention and interference. Considerable insight into database design has been obtained from performance statistics provided by the analytical model. Relative storage structure performance is shown to be highly dependent upon the types of data processing activities required. Less significantly, it is also dependent on how the database is allowed to grow over time and on the level of secondary storage contention.