Answers are grounded in this paper and cite their sources.
Much work in the analysis of proxy caching has focused on high-level metrics such as hit rates, and has approximated actual reference patterns by ignoring exceptional cases such as connection aborts. Several of these low-level details have a strong impact on performance, particularly in heterogeneous bandwidth environments such as modem pools connected to faster networks. Trace-driven simulation of the modem pool of a large ISP suggests that "cookies" dramatically affect the cachability of resources; wasted bandwidth due to aborted connections can more than offset the savings from cached documents; and using a proxy to keep from repeatedly opening new TCP connections can reduce latency more than simply caching data.
No takes yet. Share an insight, caveat, or question.
Cáceres et al. (1998) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: