Adaptive Scheduling with Client Resources to Improve WWW ServerScalability

WWW-based Internet information service has grown enormously during last fewyears, and major performance bottlenecks have been caused by WWW server andInternet bandwidth inadequacies. Augmenting the server with multi-processorsupport and shifting computation to client-site machines can substantiallyimprove the system response time and for some applications, it may also reducenetwork bandwidth requirement. In this paper, we model client-serverpartitionable WWW applications and propose adaptive scheduling techniques thatoptimize the use of client-server resource by predicting the aggregate impactof I/O, CPU and network capabilities. We present a software system calledSWEB++ which implements and supports the use of our scheduling strategies whenprogramming WWW applications. We also provide a performance-analysis frameworkbased on homogeneous client-server assumptions to identify the impact of systemloads and network bandwidth and demonstrate the effectiveness of our schedulingstrategies. Finally we present several experimental results to examine thesystem performance and verify the usefulness of the analytic model.