Application of Queueing Network Models in the Performance Evaluation of Database Designs

In this paper, we model database designs using queueing networks, giving visibility to the dynamic behaviour of the database design and allowing the database designer to experiment with different design decisions. Our approach is to abstract away the more detailed levels of the database system design by concentrating on the information that is available to the database designer at design time. It differs from other methods of database system performance evaluation in that the performance assessment is specifically targeted at the database design, not at the database system software architecture. We present a bottleneck evaluation of the Transaction Processing Performance Council TPC-C benchmark under different workload conditions and demonstrate how this affects database design decisions.