Peer-to-Peer Web Caching: Hype or Reality?

Yonggen Mao, Zhaoming Zhu, and weisong Shi
Proceedings of the 10th IEEE International Conference on Network Protocols (ICNP'02)

Overview

The design space of a peer-to-peer web caching is three dimensional: cache algorithm, document lookup algorithm and peer granularity. This paper evaluates every design combination. Two caching algorithms are evaluated: URL-based caching and content-based caching. Three document lookup algorithms: home1 (centralized indexing server), home2 (DHT), geographic-based (only query geographically close hosts by using a multicast). Peer granularity: host, organization, building and centralized web caching.

The design possibilities are evaluated using a trace. Using content-based caching combined with a DHT and host as peer granularity achieves the highest hit ratio and byte hit ratio of respectively 59.08% and 30.74%. Using a simple linear latency estimation model the average latency and latency reduction is estimated. Content based caching with geographic-based document lookup achieves the smallest latencies. For 98.8% of the requests the latency is improved.