When does a hit = a miss?

Full Paper (12 pages)
Postscript (1.5 MB) PDF (432KB)
Brian D. Davison, Chandrasekar Krishnan, and Baoning Wu

Abstract
The Simultaneous Proxy Evaluation (SPE) architecture allows one to evaluate implemented proxies in a live network environment and be able to make performance comparisons between black-box systems. When using a real workload on a live network with real content, SPE is additionally able to evaluate prefetching proxies.

The SPE architecture requires, as a component, a proxy cache that serves all cache hits with the same response time as misses. The incorporation of a proxy cache that does not reduce end-user retrieval latencies hides the presence of the cache. By slowing down the response times of cache hits to that of cache misses, a client will not be able to distinguish between the two. Squid is modified to perform in this fashion. This paper describes the relevant issues, the implementation, and the successful evaluation of this modified Squid proxy cache.

In Proceedings of the 7th International Workshop on Web Content Caching and Distribution (WCW-2002) , Boulder, CO, pages 91-102, August 14-16, 2002.

Back to Brian Davison's publications


Last modified: 20 August 2002
Brian D. Davison