Performance Evaluation of Active Database Management Systems Using the BEAST Benchmark

This paper presents the first comparative performance study of object-oriented active data- base management systems by using the BEAST benchmark. BEAST stresses the perfor- mance-critical components of active systems: event detection, event composition, rule retrieval, and rule firing. For event detection both method invocation events and transaction- al events are taken into account; this also shows some performance contributions of the pas- sive part of an ADBMS. Four systems, namely ACOOD, Ode, REACH, and SAMOS, have passed the benchmark tests of BEAST. The interpretation of the performance measurements shows several achievements in the area of active database technology, but also indicates tradeoffs (e.g. between performance and functionality). Finally, it helps to identify possible optimizations and open issues in designing and implementing active database systems.