Research On Network Optimization Algorithm Based On PTN

With the vigorous development of network technology,the number of users is increasing day by day.As a result,the packet transport network(PTN)is facing a development bottleneck,and a series of prominent problems have emerged.Layers of stacked solutions have increased the complexity of the network.In order to solve the problem of network optimization,this thesis first introduces the threetier architecture of software defined network into the PTN to optimize the complex and bloated PTN network architecture and other issues,making the original distributed control become centralized.Secondly,based on this innovative architecture,PTN optimization rules and strategies have been formulated.Then,it conducts algorithm research on the four major function modules in the controller,including topology discovery algorithm,optimal path algorithm,traffic prediction algorithm and load balancing algorithm.To improve the ability of the controller to discover the network topology,an improved topology discovery algorithm is designed in the topology discovery module to optimize the number of packets sent by the controller,thereby reducing network overhead.Aiming at the logical co-routing problem,a mathematical model of the multi-constrained optimal path is constructed,and an improved ant colony algorithm is put forward.The algorithm can improve the efficiency and accuracy in searching for the optimal path in the network.To improve the prediction accuracy of the model,the traffic prediction module needs to preprocess the obtained data at first,and then perform cluster analysis of the data.After the cluster analysis results are obtained,the model training is performed according to different clusters.The load balancing algorithm learns the load status of the link based on the result of traffic prediction,and then constructs a multi-objective function of migration cost and balance degree,and uses the optimal solution obtained by the multi-objective load balancing algorithm as the load balancing scheduling scheme.Finally,use Mininet software to create an experimental topology suitable for PTN,and conduct experimental tests on the algorithm designed in the controller.The test results show that the improved topology discovery algorithm enhances the network topology discovery ability;the ant colony algorithm significantly improves the search efficiency and accuracy through the improvement of the optimization rule;the prediction model accuracy after the clustering algorithm analysis meets the requirements and is above 89%;The multi-objective load balancing algorithm reduces the migration cost and effectively improves the overall load balance of the network.