An expert system approach for database application tuning

摘要

The paper describes a method and the principles relative to the tuning of database applications using an expert system. In database applications the maintenance concerns the reorganization of the database, generally decoupled from the software interventions. In our proposal we consider jointly the internal schema of the database and the application programs, i.e. the calls they issue against the underlying DBMS. There are several cases in which it is possible to reformulate a query to have it performed more efficiently by the DBMS. The paper presents first a method that allows to classify transactions on the basis of operations performed on the database, then the optimal internal schema is given for each class of database operations, and finally it is shown how to transform the result of this work in the form of rules for an expert system. The method has been tested using a commercial RDBMS and expert system shell. The application program has been generated artificially in order to have a mix of transactions, difficult to be found in a single application. The results show that performing periodically the database application tuning can improve significantly the performance of the overall application.