Routing, Modulation and Spectrum Assignment using an AI based Algorithm

Abstract

There is a need of the networks which can accommodate the requirements of high data rate with low blocking probability. Optical networks are used most commonly as the backbone. They support high data rates, but need to be provisioned with low blocking probability. Due to the use of traditional Wavelength Division Multiplexing (WDM), optical network are not that much bandwidth efficient as they could be. Elastic Optical Network (EON) configuration can be used to resolve this problem if Routing and Spectrum Assignment (RSA) problem is done efficiently. We are addressing the routing problem through an artificial intelligence-based algorithm. In earlier works, several optimization methods have been used for path-finding in the networks. Some of these studies has problem in terms of higher time complexity and incompleteness, due to which there is no guaranteed optimal solution. In this paper, we are using A* algorithm for routing, which is known for its best-estimated cost of a path from a source to a destination. To the best of our knowledge, this is the first attempt to study the use of A* algorithm in routing and spectrum allocation in the EON. We have modified this algorithm in a way such that, while reconstructing the path, we simultaneously check how many contiguous and continuous slots are available. So, while allocating the spectrum if the required slots are not available at that instant, we can either block or re-route the path. It saves time and decreases the blocking probability. Modified A* algorithm is expected to solve the problem only if initial seed values of the cost of the paths are not over estimated beyond a threshold. For the spectrum assignment, after evaluation of available slots while reconstructing the path, whichever slots are contiguously and continuously available, are allocated using different modulation formats. We have studied the algorithm for different network topologies, to check its validity and compared the results with the previously used algorithms.