We propose the Summary-Cache enhanced ICP, a scalable wide-area Web cache sharing protocol. Using trace-driven simulations and measurements, we demonstrate the benefits of Web proxy cache sharing, illustrate the overhead of current cache sharing protocols, and show that the summary cache approach substantially reduces the overhead. We study two key aspects of this approach: the effects of delayed updates, and the succinct representation of summaries. Our solution, Bloom filter based summaries with update delay thresholds, has low demand on memory and bandwidth, and yet achieves a hit ratio similar to that of the original ICP protocol. In particular, trace-driven simulations show that, compared to ICP, the new protocol reduces the number of inter-proxy protocol messages by a factor of 25 to 60, reduces the bandwidth consumption by over 50%, while incurring almost no degradation in the cache hit ratios. Simulation and analysis further demonstrate the scalability of the protocol.
We have built a prototype implementation in Squid 1.1.14. Synthetic and trace-replay experiments show that, in addition to the network traffic reduction, the new protocol reduces the CPU overhead between 75% to 95% and improves the client latency. The prototype implementation is publicly available [15].
Much future work remains. We plan to investigate the impact of the protocol on parent-child proxy cooperations, and the optimal hierarchy configuration for a given workload. We also plan to study the application of summary cache to various Web cache consistency protocols. Last, summary cache can be used in individual proxy implementation to speed up cache lookup, and we will quantify the effect through modifying a proxy implementation.