The continued growth of the World-Wide Web necessitates the use of proxy caches to reduce latency, network traffic and Web server loads. As proxy caches have finite storage capacity, key to the effectiveness of proxy caches is a replacement policy that can yield high hit ratio and byte hit ratio. Temporal locality and popularity, which are characteristics in object referencing behavior at Web proxy caches, are important predictors of the referencing behavior to be seen in the future. How the set of frequently referenced objects changed with time is useful information to adjust a reflection rate between popularity and temporal locality in an adaptive way.
This paper introduces Active set aware LFU, which incorporates popularity with temporal locality in a simple and adaptive fashion. Object life times are a key to detect active set change rate in active set aware LFU. Trace-driven simulations show that Active set aware LFU outperforms existing proxy cache replacement policies. In addition, Active set aware LFU can potentially improve the performance of main memory caching of Web objects.
