A centrality-based heuristic for network design to support availability differentiation

In today's society, communication networks are of paramount importance, and providing adequate levels of availability of the resources in a cost effective way is crucial for network managers. We consider the design of a high availability structure (a spine) in the network, so that a desired availability for the network flows may be achieved. The tackled problem involves the selection of the edges forming the spine and the selection of the enhanced availability for each of those edges, aiming at fulfilling a prespecified availability value for each flow, at minimum cost. We solve the formulated Mixed Integer Linear Problem (MILP) for small networks, which allows us to identify some characteristics of the spanning tree formed by the set of links with upgraded availability. Afterwards, using that information, we propose a heuristic based on a centrality measure, which allows us to devise the appropriate set of links, and which may be used in larger networks. Experimental results show the effectiveness of the resolution approach in finding spines equal to the optimal ones or to the best known solutions.