Adaptive Web Proxy Caching Algorithms

Web proxy caching is a well-known technique for reducing access latencies and bandwidth consumption. As in other caching systems, a replacement policy is necessary for determining when and what to evict from the cache, and many proxy caching algorithms have been proposed and evaluated over the years. Yet many of these algorithms were motivated by the Internet traffic trends in place during the design of the algorithms (1997–1999). The recent proliferation then subsequent decline in the use of peer-to-peer systems suggests that the performance of proxy caching algorithms may have changed dramatically since their inception. This paper analyzes the distribution of current web content and re-evaluates various proxy cache replacement algorithms including LFU, LRU and several GreedyDual variants. Based on the change in Internet traffic trends identified, two new web caching algorithms are proposed: a local policy that maintains a list of popular URLs and a global policy that partitions the cache into distinct regions, each of which is sized according to the popularity of the type of object being cached. Our results indicate that for small cache sizes LFU and LRU have higher byte-hit rates than several GreedyDual variants on modern Internet traffic. We also found that maintaining a hot list improves byte-hit rate, while a cache-by-regions approach improves hit rate.