Scalable design method of attractors in noise-induced virtual network topology control

One approach to accommodating IP traffic on a wavelength-routed network is to construct a virtual network topology (VNT) and reconfigure the VNT according to traffic changes. Our research group has proposed a VNT control method based on attractor selection, which is a model of the behavior by which living organisms adapt to unknown changes in their surrounding environment. Guided by attractors, the VNT control method searches for a solution: a VNT that can accommodate the IP traffic. Attractors are a subset of the equilibrium points in the solution space and correspond to VNT candidates. It is crucial to design the attractors properly, since they define the attractive states in the VNT control. If the VNT candidates are not designed properly, it takes a long time for the method to find a solution. This paper therefore proposes a method for designing VNT candidates. Our basic approach is to prepare VNT candidates in which the bottleneck links (lightpaths) are different from each other. However, our exhaustive algorithm based on this approach has the problem of requiring large amounts of computational time for large-scale networks. We therefore propose a method that hierarchically contracts a network topology to enable application of our algorithm to large-scale networks. Evaluation results show that the VNT control method using attractors obtained by our method achieves shorter convergence times.