Designing backbone networks using the generalized steiner problem

In the design of metropolitan optical fiber networks, a commonly applied requirement is to ensure the existence node-disjoint-paths between pairs of distinguished nodes of the network. In this way, when a failure occurs in some component (link or node), the network will remain connected, a basic requisite for maintaining an operational state. In this work, we introduce a Greedy Randomized Adaptive Search Procedure (GRASP) for designing a survivable backbone. The algorithm builds topologies which comply with heterogeneous node-connectivity requirements. The method is applied to a set of problem instances with different connectivity requirements, obtaining results which appear promising.