A new cache approach based on graph for Web servers

This work deals with the problem of data cache for dynamic Web servers. A set of requests (which number could be large) is submitted to a server for computing. Every request is modelled by a series-parallel graph of elementary tasks. These tasks are computed and their results are submitted to be stored into a cache. Due to the limitation of the cache capacity, some strategies should be used to select the right ones. The aim of This work is to determine, in a static session, what tasks results have to be cached in order to reduce the computing time. The problem has been formalized using graphs representation whose complexity is briefly shown to be NP-hard. Then, heuristics are proposed to select the cached data. Finally some preliminary experimental results are discussed.