Research On Intelligent Routing Optimization Technology For SDN

To support the rapid development of cloud computing,the Internet of Things and other technologies,numerous data centers have been built.As the amount of service data carried by data centers increases rapidly,the network architecture is flexible and effective,and can adapt to various types of large-scale data transmission.Contrary to traditional network architecture,SDN architecture not only enables centralized control and open network programming,but also separates data plane and control plane.As a new network architecture,Software defined Network(SDN)provides a new way to solve the traditional routing problem.Aiming at the low resource utilization of data center network,this paper proposes an intelligent routing algorithm based on deep reinforcement learning on the basis of the research of intelligent routing optimization technology for SDN.The algorithm on the SDN architecture add a plane with depth of reinforcement learning knowledge,knowledge of the plane DQN agent for SDN architecture provides the network state information,namely the link bandwidth available,link delay and link information such as packet loss rate and calculate the best route,using the modified greedy -greedy function,implements a intelligent routing algorithm.Then the parameter selection of the algorithm in this paper is determined by experimental comparison.Experimental results show that the proposed intelligent routing algorithm based on deep reinforcement learning is superior to Dijkstra algorithm and Q-Routing algorithm in terms of packet loss rate and delay.Graph neural network can extract information from graph structure data and is widely used in many fields.This paper presents an intelligent routing algorithm based on graph neural network.This algorithm combines the topological information perception ability of graph neural network with the self-training ability of deep reinforcement learning to improve the intelligence of network routing strategy.GNN constructs the network topology as a graph,in which the link of the network topology is the entity of the graph.According to the graph structure,an iterative messaging algorithm is run between the hidden states of the link.The deep reinforcement learning agent selects the optimal path according to the results of message passing,and the deep learning framework uses the improved precedence experience replay mechanism to train GNN and realize the intelligent routing algorithm.The parameter selection of the algorithm in this paper is determined by experimental comparison.Finally,the simulation results show that the proposed intelligent routing algorithm based on graph neural network performs well and is more robust to changes in graph structure of the network,such as adding or deleting links.