The proposed research will develop a new generation of queueing networks that can model new features not previously representable, e.g. bursty, correlated traffic, data losses and load balancing. It will follow an approach that
Validation, to some predefined level of precision, will be against existing exact models, where their assumptions hold, and against simulation models where they do not hold. Results will relate to both resource-based measures such as throughput and mean queue lengths, as well as user-oriented measures, typified by response time distribution. If successful, the new approach would revolutionise performance modelling based on queueing network analysis: the range of systems modelable would be vastly expanded and the efficiency of models' solutions (in terms of both execution time and storage requirement) would be enhanced dramatically by their compositional nature.