A replicated resource architecture for high performance network service

Distributed Learning Environments represent the hope that communications and information technology can improve and widen access to education while maintaining and improving its quality. Such environments consist of network applications and services. Good interactive response time is crucial to their success. Slow responses can quickly dissuade teachers and learners alike from investing their time in the use of these services. Responsiveness timings taken across 155Mb/s IP/ATM networks have exposed traditional monolithic server performance as the main bottleneck in interactive response time. A strategy of providing bigger and faster monolithic server hardware in response to each occurrence of system slow down is not a good solution as it is expensive and inflexible. Cluster computing has proven a successful and cost effective alternative to conventional supercomputing and it would now seem to be appropriate to investigate its application to the problem of high performance network service provision. In order to research this issue a replicated resource architecture has been designed to harness the combined power of multiple independent computers. The architecture is outlined and an initial implementation of its core component, a coherence server, is described. Results are presented which indicate that this approach is viable within the context of Distributed Learning Environments.