Server Location Method for Keeping Shorter Distance from Users to Servers During Failures

Recently, large contents in the Internet have increased loads of contents servers, networks and data centers, which may degrade the quality of services. To overcome this problem, some mirror servers providing the same contents are located on a network and a request is navigated to one of the mirror servers. A user must access to one of the servers and the hop length of a path from a user to a server should be short even during link failure. As it affects the performance, the location of the mirror servers is important. In this paper, we address the server location problem, which determines the location of the servers satisfying the following the constraint: any users can access servers within a small hop count even if some links fail. First, we formulate this problem and prove that it is NP-hard. Second, we present a heuristic polynomial-time algorithm for the case that the increase of the distance between a user node and servers during failure is restricted to be the number of nodes. Furthermore, we also present an approximation algorithm to solve this problem when the number of servers to be accessed is restricted to one and the number of simultaneously failed links is a constant positive integer, and we evaluate the performance by using some actual network topologies.