Performance Evaluations of Replacement Algorithms in Hierarchical Web Caching

Web caching plays an important role in many network services. Utilization of the cache in each level (server, proxy, and client) of network forms a web caching hierarchy. A major problem of the hierarchical caching is the poor performance due to the interference between upper-level and lower-level caches. This paper investigates the replacement algorithms applied in the network caching hierarchy through trace-driven simulation experiments to identify the performance bottleneck. Our investigation focuses on three fundamental replacement algorithms: LRU, LFU and SIZE, because many other replacement algorithms are mainly the variations and/or combinations of the three fundamental algorithms. Through extensive experiments, we have acquired useful performance features of these algorithms and their combinations at different levels. Thus, our work may serve as a reference for the design and deployment of efficient replacement algorithms in the caching hierarchy for web engineering.