The Study Of Network Routing Algorithm With Node Representation

With the development of the mobile Internet,the explosive growth of the number of users and multimedia traffic are bringing challenges to network routing technology.Nowadays,the transition of network state is more complex than ever before.However,traditional rule-based routing methods cannot learn potential valuable information from massive network flows and network status.Meanwhile,dynamic rule is not available in these methods.The emergence of the software defined network(SDN),which decouples the data plane and control plane of the router and gives the network device programmable flexibility,makes dynamical modification of routing rules available.However,SDN couldn't make automatic network analysis and deal with problems without human intervention.Therefore,the industry has proposed the idea that applies machine learning,deep learning and other technologies to the control plane of SDN.Not lirmited to SDN networks,such emerging technologies have made good performances in the network field,such as network data analysis,data transmission and other issues.This paper introduces machine learning into network routing.According to the common idea in these technologies that encode vectors to represent information,a new network device representation-node vector as well as correlation theory is proposed.Since the network device can also be abstracted as a node,node vector also named Node Representation(NR).Around node vectors,the main contributions of this paper are as follows:First,we proposed a heuristic network routing algorithm in order to solving basic routing problem.In this algorithm,machine learning is used to calculate node vectors that encodes distance information into node vectors.According to these vectors and routing principle,a basic routing algorithm is proposed.Second,as an object of network structure,node vectors provide basic features for the application of learning-based methods in network.Based on these basic characteristics,neural networks are introduced into the network for QoS constrained routing and network state analysis.Third,we considered different application scenarios of this method,and potential applications.Simulation and analysis on several scenarios proved that the proposed method is reliable and stable in routing tasks.