Optimizing the Evaluation of Calculus Expressions in a Relational Database System

Abstract

There is an interesting search strategy (due to James B. Rothnie) for efficient implementation of a limited kind of selection criterion for a relational database. This strategy is here generalized to arbitrary relational calculus expressions, and an analysis of the resulting improvement of performance is given. The strategy is used in a relational database system TGR and an overview of the architecture of this system is presented. TGR uses microprogrammed database primitives for searching the database. This approach is very similar to the use of a database processor although it also allows flexible change of processor design. The behaviour of TGR in evaluating typical queries is analyzed and the results are used for pointing out the bottlenecks in a relational database system with a particular type of structure. As a conclusion the construction of a database processor with the database primitives from TGR as instruction set is recommended. This would be a step towards getting acceptable performance in a relational database system.