Cache Modelling in a Performance Evaluator of Parallel Database Systems
Shi Zhou, Neven Tomov, Albert Burger, Hamish Taylor · 1997
Cache modelling is an important issue in developing an analytical performance evaluator to estimate performance for applications running on parallel DBMSs. This paper describes a cache model developed for parallel cache management in Oracle7 Parallel Server. Some preliminary results have also been obtained by using the cache model to predict cache hit ratios for varying database size and varying number of participating processing elements. 1. Introduction The performance of a parallel database management system (DBMS) depends on many factors, such as the characteristics of the DBMS, the architecture of the system, the pattern of queries, the way in which queries are optimised and the way in which items of data are distributed across the nodes of the system (data placement). Prediction of the effect on performance of changes to various system parameters relies heavily on having an accurate model of performance measures such as throughput, response time and utilisation for given system ...