To provide performance models of networks of computing resources;
To investigate global state-dependence, e.g. blocking and adaptive routing;
To develop performance models of interconnection networks;
To represent parallel workloads and cache coherency overheads;
To investigate load balancing and scheduling issues.
Progress has continued to focus significantly on response time distributions in queuing nodes and networks of switches [2,3,5]. In addition, the work on queuing networks with negative customers has advanced substantially, with new results on response times as well as new applications, in particular in the area of reliability which can be used to model data loss in networks [2,3]. The work on blocking in Banyan networks has also advanced . Work on modelling parallel workloads has been concentrated on developing a formalism capable of expressing the behaviour of both software and hardware systems , together with specific analyses of the behaviour of applications in a cache coherency context . A paper on cache coherency in an SCI-based parallel system has been submitted for external publication . It is intended to concentrate on scheduling and load balancing issues in the final year of the project.
 A J Field, P G Harrison; 'Performance analysis of parallel architectures using the SCI interconnect', Research Report, submitted for publication, 1995.
 P G Harrison, E Pitel; 'Response time distributions in tandem G-networks', Journal of Applied Probability, 1995.
 P G Harrison, E Pitel; 'Modelling queues with breakdowns via negative customers', Research Report, submitted for publication, 1995.
 P G Harrison, B Strulo; 'Process Algebra for Discrete Event Simulation', In Proc. 2nd QMIPS Workshop, eds N Gotz, U Herzog, M Rettelbach, University of Erlangen, 1993.
 A de C Pinto, P G Harrison; 'An approximation for end-to-end delay distributions in buffered multistage interconnection networks', In Proc. Int Conf on Local and Metropolitan Communication Systems, Kyoto, December 1994.