We will develop stochastic models of internet systems using a compositional approach. This will include: Abstract modelling of resources and their configuration, workload modelling, the development of submodels corresponding to internet sybsystems and submodel composition. Workload models will be derived by analysis of traffic on local systems at Imperial college and the London Metropolitan Area Network (MAN), and additional data supplied through Metron Technology Ltd.
New techniques will be explored for:
Performance will be predicted in terms of both resouce-based measures, such as utilisation, throughput and mean queue lengths, as well as response time averages and distributions. Models will be implemented initially in Mathematics and validated with respect to simulation and measurement on real systems including the "Kensington" data mining application at Imperial College. Validated models will be used to compare the efficiency of alternate internet and algorithm configuration for typical workloads and traffic patterns.
All data collected for the QUAINT project is available here: <http://www.doc.ic.ac.uk/~uh/QUAINT/data/> .