Performance models of a firm's proxy cache server

We examine the impact of installing a proxy cache server (PCS) on overall response time to Web requests. We analyze how various factors affect the performance of that server. Our research specifically identifies a "crossover probability", the minimum cache "hit rate" probability at which installing a PCS becomes beneficial. We find that this probability decreases as the arrival rate of Web requests or the average file size increases. In particular, the benefits of installing a PCS are more pronounced when the firm's users exhibit heavy Web accesses. We also find a "diminishing rate of return" phenomenon in terms of enhancing the PCS's performance. The managerial implication is that it may not pay to choose an overpowerful PCS as the marginal reduction in the overall response time becomes unjustified. Moreover, the "bottleneck" effect of the firm's network bandwidth is investigated and demonstrated. (C) 2000 Elsevier Science B.V. All rights reserved.