Dimensioning Transport Networks for VPNs over Capacities with Stepwise Costs

The virtual private network (VPN) services have become widespread and many of research activities aim to provide efficient methods for VPN design. However, only a few discrete bandwidths at different prices are available resulting in stepwise capacity cost functions (CCFs). Although numerous general heuristics are proposed for VPN design, they cannot deal with stepwise CCFs. This paper proposes a three step heuristic: first it "smooths" the stepwise CCFs, then it solves the modified problem using simulated allocation and finally it rounds either up or down the capacity steps via rerouting some of the traffic demands. As a reference we present the integer linear program (ILP) formulation of the considered problem and we also use a three-phase heuristic where we first smooth the stepwise CCF, then we configure the VPN (by e.g., simulated allocation) and then we round up the capacity values. We conduct simulations to evaluate the proposed methods and we show that enabling rerouting instead of always rounding up significantly increases performance. In the last section we illustrate the applicability of the method on a case study