Caching Strategies for Data-Intensive Web Sites
Résumé
A data-intensive Web site is a Web server that accesses large numbers of pages whose content is dynamically extracted from a database. In this context, returning a Web page may require costly interaction with the database system (for connection and querying) thereby increasing much the response time. In this paper, we address this performance problem. Our approach relies on the declarative specification of the Web site. We propose a customizable cache system architecture and its implementation, in the context of the Weave Web site management system developed at Inria. The system can cache database data, XML fragments or HTML files. We illustrate it using a Web site derived from the TPC/D benchmark database. Based on experiments using our test platform WeaveBench, we assess the performance of various caching strategies. The results clearly show that a mixed strategy is generally close to optimal (static files). Finally, we derive guidelines for specifying cache management based on the Web site specifications.