Research On Routing Spectrum Allocation Algorithm In Elastic Optical Network

With the rapid growth of traffic in today's society,the original optical network based on wavelength division multiplexing due to its shortcomings of large spectrum granularity caused a lot of waste of spectrum resources,making it difficult to meet today's traffic demands,so the elastic optical network It came into being.In the elastic optical network,the most important research problem is the problem of routing routing allocation(Routing Spectrum Allocation,RSA).This paper mainly studies the RSA algorithm in the elastic optical network.Aiming at the problem of spectrum fragmentation in the network,a free frequency slot value measurement method based on the access blocking rate is proposed,and on this basis,two RSA algorithms PABR are proposed.PABR-RSA(RSA Algorithm based on Posterior Access Blocking Rate)algorithm and RL-RSA(RSA Algorithm based on Reinforcement Learning)algorithm.The main research work of this article is as follows:(1)Aiming at the problem of spectrum fragmentation in the RSA process,a method for considering the value of the free frequency slot of spectrum fragment is proposed.This method is based on the blocking probability of a certain frequency slot on the service access link,combined with the probability distribution of the bandwidth demand of the service in the traffic model,and quantifies the value of the idle frequency slot in the link(2)Analyze the causes of spectrum fragmentation in the network,and propose the PABR-RSA algorithm.Based on the conclusion in(1),this method quantifies the value of the idle frequency slot with reference to the degree of spectrum fragmentation of the link,and further quantifies the value of the network link to allocate routing spectrum resources to reduce the service blocking rate.The simulation results show that compared with the classic RSA Algorithm based on Shortest Path + First Fit(SF-RSA)based on the shortest path + first hit(SF),the PABR-RSA algorithm is different in In the case of network topology and different business volumes,the three indicators of business blocking rate,network average link value,and network link value standard deviation are superior to the benchmark algorithm.(3)Considering the problem of using a fixed RSA strategy for the heuristic RSA algorithm so that it performs poorly in complex and dynamic scenes.This paper proposes a RL-RSA algorithm based on Reinforcement Learning(RSA Algorithm based on Reinforcement Learning)that uses reinforcement learning,establishes the reinforcement learning model of the RSA process,solves the problem of uncertain action space in the RSA process,and determines the appropriate the value of the hyperparameter.The simulation results show that under different network topologies and different traffic volumes,the three indicators of service blocking rate,network average link value,and network link value standard deviation are superior to the benchmark algorithm,and the RL-RSA is verified through experiments the robustness of the algorithm.