Abstract
The goal of our research is to predict performance of applications using ORACLE RDBMS and then to propose tools to optimize performance.
The performance we are interesting in, is the performance as perceived by users. We then specially study user’s transactions response time.
Our method is based on measurement, and the first step was to measure performance on existing applications using ORACLE RDBMS. We have developed several software tools:
- user simulator
- application generator
- workload application generator
- measurement tools to measure user transactions response time and system activity during a transaction execution.
The second step consisted in data collection and data analysis of measures. The data analysis, based on statistic methods, has permitted to extract the most influent factors and to understand how they can enhance applications performance (Rumpler and Polo, 1993).
We are now able to present the most part of these results.
The last step will consist in building the rules of an expert system for configuration and tuning assistance of ORACLE RDBMS applications. We also analyse the impact of operating system (UNIX) parameters on performance, and these information will complete our expert system possibilities.
The present paper will describe precisely this research with tools developed, methods used, and results.
