SDN And Machine Learning Based Control Mechanism In Optical Network

In recent years,with the emergence of various broadband services and the development of technologies such as Internet of Things and big data,network traffic has shown a continuous growth trend.Meanwhile,the development of optical networks that serve various emerging technologies and applications has been posed an unprecedented challenge.Large capacity,high flexibility and scalability become the development direction of the future communication network;network topology is also more and more complex,in order to meet the basic needs of subscribers and guarantee the normal operation of QoS mechanisms,it is more crucial to implement congestion control and optimize the wavelength allocation mechanism on optical networks.Based on the National Natural Science Foundation of China,this paper focusing on the structure of TWDM optical networks,software defined optical network and bandwidth allocation algorithms,proposes a control mechanism based on prediction of optical network performance using machine learning.The scheme is based on software defined optical network system structure,considering the real-time status of network resources and the traditional routing resource allocation algorithm,a new dynamic bandwidth algorithm is proposed.Further improve network bandwidth utilization,reduce business blocking rate and delay.The main work and innovation are as follows:(1)The simulation platform of TWDM optical network is set up,and the key multiplexing functions such as time division and wavelength division are simulated,the key performance data collection is completed,such as the source and destination nodes of the service,the bandwidth and the duration of the request,the link bandwidth utilization during fixed time periods and so on.(2)A theoretical model based on machine learning is proposed to predict the performance of optical network.Its structure and parameters are determined through the analysis of machine learning algorithms.In view of the current increasingly serious network congestion problem,a scheme based on machine learning for predicting the congestion degree of optical networks is proposed and implemented.Simulation results show that the proposed scheme can effectively predict the congestion of the system and achieve the functions of early warning and theoretical guidance.(3)In order to further control and reduce network congestion,a resource allocation scheme based on bandwidth utilization prediction in SDON is proposed.The simulation results show that the scheme can effectively reduce the traffic blocking rate and delay and improve the system bandwidth utilization efficiency by centralized routing and resource allocation through the SDN controller according to the real-time link status.