An Optimization Of Rmsa Based On The Genetic Algorithm For Unicast/Multicast Requests In EONs

Orthogonal Frequency Division Multiplexing, also known as OFDM, is recently well studied as a key modulation technology in Elastic Optical Networks (EONs) due to the high efficiency and flexibility of resource allocation and the ability of impairment tolerance. For the pending network requests, the modulation level can be computed after the routing information is determined and then we will flexibly assign the requests with a certain number of sub-carriers. The above procedure including routing, modulation and spectrum assignment is called RMS A in the following paragraphs which is exactly our main research aspect. There are already several studies for the RMSA problem and also some algorithms are provided to resolve it. Besides, the RMSA problem turns out to be a NP-complete problem. During our research, we divide the problem into two sub-problems, the first one is the routing and modulation level problem and the second one is the spectrum assignment problem. We will resolve the two sub-problems in sequence. In this article, we develop a heuristic algorithm called genetic algorithm (GA) to study the RMSA problem which can provide an effective method for the spectrum assignment of the optical network requests. The GA is designed for the multi-objective problem and in the scenario when the network traffic is low (which means the resource is adequate for all the requests and none will be blocked), it tries to minimize the required maximum slot index among all the fibers in the topology. On the contrary, when the traffic climbs and the resource is insufficient, the blockings occurs and the GA tries the best to serve all the requests successfully and the objective then is to minimize the blocking probability of the requests.In this article, we simulated our GA to solve the dynamic RMSA problem on both the NSFNET topology and the US Backbone topology. The results tell that firstly when there is only unicast requests, our proposed GA can provide a better result on the load balancing and the blocking probability than several existed algorithms. And secondly when the multicast requests are considered, the GA is also superior to the existed heuristic algorithms which based on the SPT and MST for multicast dynamic RMSA problem. At last, we also checked the convergence of the GA and the results show that it can converge within10and20generations under low traffic and high traffic scenario respectively.