Benchmarking Database Systems in Multiple Backend Configurations

The aim of this performance evaluation is twofold: (1) to devise benchmarking strategies for and apply benchmarking methodologies to the measurement of a prototyped database system in multiple backend configurations, and (2) to verify the performance claims as projected or predicted by the designer and implementor of the multi-backend database system known as MBDS. Despite the limitation of the backend hardware, the benchmarking experiments have proceeded well, producing startling results and good insights. By collecting macroscopic data such as the response time of the request, the external performance measurements of MBDS have been conducted. By collecting microscopic data such as the time entering and leaving a system process, the internal performance measurements of MBDS have been carried out. Methodologies for constructing test databases, directories, and requests have been devised and utilized. The performance evaluation studies verify that (a) when the database remains the same the response time of a request can be reduced to nearly half, if the number of backends and their disks is doubled; (b) when the response set of a request doubles, the response time of the query remains nearly constant, if the number of backends and their disks is doubled. These were the performance claims of MBDS as predicted by its designer and implementor.Chief of Naval Researc