Placing bypass capable nodes in two-layer networks

In two-layer networks there are a lot of ways to improve virtual and physical topology and capacity allocations in terms of cost. Because switching costs are usually higher in the upper layer, the introduction of bypass links, bypassing the expensive upper layer interfaces, can be beneficial. In the paper the authors discuss the placement of nodes with bypass functionalities. The authors use a physical network topology with 50 nodes based on the Germany50 reference network. The authors' approach for finding the bypass links and the corresponding bypass nodes is based on a mixed integer programming approach with a simple cost model for switching in both layers. The starting point is a network without bypass functionality. This is basically a one-layer network, as all traffic is switched in the upper layer. In a first step the authors find the optimum capacity allocations in this network, taking the interface granularities into account. After this the authors calculate the resulting transit traffic in all nodes. If the transit traffic between two next nearest neighbour nodes is larger than a certain threshold, a bypass link between them is introduced between them. The new virtual (upper layer) topology is then used to optimize the network again. This procedure leads to considerable cost reductions, but is computationally much simpler than a full two layer optimisation including all possible bypass links.