Locating concentrators in centralized computer networks

Topological design of centralized computer communication networks is a complex problem that is generally solved in two phases. The first phase of the design process involves dividing network nodes (terminals or clusters of terminals) into groups, and selecting a concentrator location for each group so that all the nodes in a group are assigned to the same concentrator. The next phase determines topology of links that connect network nodes to concentrators and concentrators to each other and to the central computer. The design problem studied in this paper contains some aspects of both phases. In this problem locations of concentrators, assignments of user nodes to concentrators and the topology of the links connecting concentrators to the central computer are jointly determined. The proposed design method is built around the well known sweep heuristic which is used to partition the node space into sectors. Each of these sectors contain a ‘backbone’ path connecting concentrators to the central computer.