On prefetching in hierarchical caching systems

Y. Thomas Hou, Jianping Pan, Chonggang Wang, Bo Li

Abstract: Hierarchical caching is deployed to scale up the explosive Web growth, and the expiration-based mechanism is adopted as an economic means to support the weak consistency in this context. However, given a hierarchy, the user perceived performance heavily depends on its position. Normally, a user near the hierarchy leaf suffers higher miss rate and longer response time. Such an intrinsic property can discourage users from participating in any hierarchical caching systems. In this paper, we analyze the performance of a proposed approach, i.e., freshness and retrieval threshold based cache prefetching, to mitigate the bias against leaf users. We also use ns-2 to further substantiate our analysis. By adopting this approach with the appropriate parameters, the fairness among users within a caching hierarchy can be considerably improved.