Research On Routing And Resource Assignment Algorithm Based On Machine Learning

In the 21 st century,cloud computing,big data and other technologies are booming.At the same time,a large number of users are pouring into the Internet,and the optical communication networks face new challenges.To adapt to the increasing surge of user data,the transmission technology in optical network has developed from wavelength division multiplexing technology to elastic optical network technology,which increases the transmission capacity of the network and effectively alleviates the insufficient transmission capacity of the network.This makes the optical communication network can carry more user demand data and improve the comprehensive performance of the optical network.At the same time,theuse of multi-band transmission demands can significantly increase the network transmission capacity without modifying the existing optical fiber infrastructure.In order to provide users with better transmission quality and service quality of the optical network environment,this paper focuses on the issues related to routing and resource allocation policies in elastic optical network.Firstly,this paper considers the spectrum fragmentation problem in the elastic optical network,and proposes the routing and resource assignment policy of reinforcement learning based on deep deterministic policy gradient.This policy to return specific rewards by the spectrum fragmentation measure.After simulation,this policy,compared with the traditional policy,can reduce the spectrum fragmentation rate in the network by about 10%.At the same time,this policy also reduces the demand block in the optical network,and increases the overall demand capacity in the optical network.Secondly,in multi-band elastic optical networks,user demands need to be carried on hardware resources such as nodes and links.The user demand cannot be transmitted if these hardware resources not work,and the user demand would be interrupted.In order to reduce the possibility of demand interruption in multi-band elastic optical networks,a policy for optimizing demand interruption based on deep deterministic policy gradient is proposed.Through simulation and result analysis,the policy,which proposed in this paper,can reduce the occurrence of demand interruption in multi-band elastic optical networks,compared with the traditional strategy,by about 1.5%.And,in terms of network spectrum fragmentation,reduced by about 7%.