Class-based Cache Management for Dynamic Web Content

Caching dynamic pages at a server site is beneficial in reducingserver resource demands and it also helps dynamic page caching atproxy sites. Previous work has used fine-grain dependence graphsamong individual dynamic pages and underlying data sets to enforceresult consistency. Such an approach can be cumbersome or inefficientin dealing with an arbitrarily large number of dynamic pages. Thispaper studies partitioning dynamic pages into classes based on URLpatterns and the proposed scheme allows an application to specifypage identification, cachability and data dependence and invokeinvalidation for a class of dynamic pages. To make this schemetime-efficient with small space requirement, lazy invalidation is usedto minimize slow disk accesses when identifications of dynamic pagesare stored in memory with a digest format. Selective precomputing isfurther proposed to regenerate stale pages and smoothen load peaks. A datastructure is developed for efficient URL class searching during lazyor eager invalidation. This paper also presents design andimplementation of a caching software called Cachuma which integratesthe above techniques, runs in tandem with standard Web servers, and allowsWeb sites to add dynamic page caching capability with minimal changes.The experimental results show that the proposed techniques canefficiently handle class-based page invalidation and areeffective in reducing server response times for tested applications.