Benchmarking the join operations of relational database machines.

Abstract

Over the past several years benchmarking has been developed into an effective technique for performance analyses of computer systems. Relational database machines are relatively new compter systems for which a benchmarking technique does not yet exist;. The benchmarking of relational database machines involves the indent ification and design of test programs through which relevant performance data can be gathered and interpreted. All features of relational database management must be considered when designing these test programs. The join operations are an important feature of relational database management. The test programs for the join operations necessarily include the repetition of certain queries during which specific join parameters are varied. These parameters include: tuple size, relation size, disk placement, and the use cf indices. A number of join operations have been benchmarked. These operations are equality joins, inequality joins, three-way joins, and virtual joins (i.e., views) . In addition, a number of relational database machine configurations have been utilized for benchmarking the join operations. The highlights of the thesis can be found in its contribution tc a benchmarking technique for the join operations and its conclusions on the performance analyses of various relational machines in operating joins.