Resource Management for Delivery of Dynamic Information

Abstract

Information delivery via the web has become very popular. Along with a growing user population, systems increasingly are supporting content that changes frequently, personalised information, and differentiation and choice. This thesis is concerned with the design and evaluation of resource management strategies for such systems. An architecture that provides scalability through caching proxies is considered. When a cached page is updated at the server, the cached copy may become stale if the server is not able to transmit the update to the proxies immediately. From the perspective of the server, resources are required to transmit updates for cached pages and to process requests for pages that are not cached. Analytic results on how the available resources should be managed in order to minimise staleness-related cost are presented. An efficient algorithm that the server can use to determine the set of pages that should be cached and a policy for transmitting updates for these pages are also presented. We then apply these results to page fragments, a technique that can provide increased efficiency for delivery of personalised pages.