Benchmarking Parallel SQL Database Machines

New parallel database systems are being developed to support the data processing needs of large commercial organisations: these offer increased performance in terms of processing throughput and enhanced support for both on-line transaction processing and management information provision. In order to evaluate the performance of these systems a new SQL benchmark has been developed which models the requirements for real database applications. Based on a large commercial database the benchmark system includes the specification of the test or “mimic” database and sets of transactions and queries, and the data/transaction generation tools. The benchmark uses synthetic data which closely resembles real world data and provides for evaluation of the scalability of system under test.