In this paper a two-level hierarchical model for the location of concentrators and routers in computers networks is presented. Given a set of candidate locations and the capacities of concentrators and routers the problem is to determine how many concentrators and routers should be used, where they should be located and which users to assign to each concentrator and which concentrators to allocate to each router.A Lagrangean relaxation is used to provide lower bounds for this problem, which are tentatively strengthened by incomplete optimization through the use of CPLEX. A tabu search meta-heuristic is then developed. Computational experiments are performed using a set of randomly generated problems. 2006 Elsevier Ltd. All rights reserved.