Performance prototyping of data management applications

Abstract

This paper describes the construction of a performance model or prototype of a proposed data base management system (DBMS) application by the application of certain techniques from the performance measurement and data base management disciplines. Performance measurement is conducted by using a “drive” workload that initiates the “real” workload with reasonable fidelity. The “drive” workload is constructed in terms of data base variable primitives, which are “inverted” to performance variable primitives in order to construct a resource-based drive workload. The application of this drive workload to a simulation model is illustrated, showing the use of the methodology to “tune” important DBMS, Operating System and application program parameters.