Capacity Planning of a Wide-Sense Non-Blocking Generalized Survivable Network

Generalized Survivable Networks (GSNs) have two interesting properties that are essential attributes for the future backbone networks – full survivability against link failures and support for dynamic traffic demands. GSN incorporates the non-blocking network concept into the survivable network models. Given a set of nodes and a topology that is at least 2-edge connected, a certain minimum capacity is required for each edge to form a GSN. The edge capacity is bounded because each node has an I/O capacity limit which serves as constraints for any allowable traffic demand matrix. The GSN capacity planning problem is NP-hard. We first give a rigorous mathematical framework, then we offer two different solution approaches. The two-phase approach is fast but the joint optimization approach gives a better bound. We carried out numerical computations for eight networks with different topologies and found that the cost of a GSN is only a fraction (from 52% to 89%) more than a static survivable network.