Hierarchical performance analysis models for data base systems

Abstract

This paper presents a comprehensive set of hierarchically organized synthetic models developed for the performance evaluation of data base management systems. The first set of algebraic models for data base management system itself contains a program behavior model, a retrieval model, a logical data base model, a physical data base model, and a data processing model. These models are intended to clarify logical and physical data base structures and essential operations in the data processing on them. Another set of models for performance analysis contains a macroscopic program behavior model, a storage model, a processor model, a user behavior model, and an interactive model. In this set this paper is particularly concerned with the first 3 models. The macroscopic program behavior model is based on the discussion of data locality and allows us to estimate the frequency of data page loading expected on a given application program and a data base. The storage model enables us to estimate the traverse time of data page loading. Finally the processor model allows us to evaluate the data retrieval processing time of data manipulation commands of a given data base structure, a data base management system and a set of application programs under the multiprogramming environment. These models are to be applied to the optimization of the application programs or the data base structure in order to obtain a higher data retrieval performance.