- In this paper, web proxy servers and mirror sites that cache either partial or complete information of their parent web servers are considered. These proxy servers are usually located strategically, for example near high-user-demand locations. Using proxy servers would significantly reduce latency for the users in retrieving information as well as alleviate congestion in the network. The main problem addressed in this paper is determining the optimal number and locations of proxy servers in a network to minimize costs subject to delay, throughput and demand constraints. For a given set of proxy server locations, it is assumed that client or user requests at a location will always be sent to the nearest server. Thereby each client-server system can be modeled as an independent queueing network for which performance measures such as delay distribution and throughput are obtained. These performance measures are used in an optimization problem that is formulated to determine the optimal number and optimal location of proxy servers. A heuristic called the DEJAVU algorithm is developed to solve the optimization problem. Based on a comparison with a genetic algorithm, it can be concluded that the DEJAVU algorithm produces near-optimal to optimal results in a very short time. 2002 Elsevier Science B.V. All rights reserved.