A performance investigation of web caching architectures

Web caching has been recognized as an effective approach to improve performance by storing frequently-accessed documents from originating servers at storage sites "closer" to requesting clients in order to reduce the load on the network bandwidth which in turn can reduce user response time. In this paper, we examine the performance of various architectures for cache cooperation in a web environment. The architectures under consideration include: distributed, hierarchical, and a hybrid of the two. A hierarchical web caching architecture utilizes a series of caches placed at different levels of the network (i.e. client, institutional, regional, and national) and user requests are satisfied at the lowest level possible (the first one that stores the desired document). Distributed web caching incorporates multiple cooperating caches all at the same level. The hybrid approach chosen in this paper contains a hierarchy of distributed caches in the hopes of exploiting advantages from both distributed and hierarchical caching. Using simulation as the tool, we compare the performance (mean response time, overall cache hit rates) of these architectures under a variety of different scenarios in an attempt to identify the conditions under which each is beneficial.