Design And Implementation Of SDN Network QoS Optimization Based On ONOS

With the rapid development of network technology,the expansion of network scale and the increase in the number of applications,users have raised requirements for the guarantee of network service quality,and there is an urgent need for efficient routing algorithms to guarantee the QoS requirements of traffic flows.However,the traditional network architecture is complex,and it is difficult to obtain a global view,which limits the design and deployment of routing algorithms and cannot provide ideal QoS guarantee.The proposal of the SDN architecture decouples the control plane and the data plane,and brings new ideas for QoS routing.Based on the SDN architecture,the routing algorithm can be implemented on the control plane according to the QoS policy,and the flow table can be installed on the data plane by the Open Flow protocol.At the same time,the popularity of machine learning algorithms has also brought new research directions for QoS routing optimization in SDN networks.In this paper,ONOS controller is selected as a research tool,combined with machine learning algorithms,to study routing schemes that provide QoS assurance for traffic flows in SDN networks.The main work includes the following three parts:First,a QoS routing scheme based on utility value and traffic prediction is designed.By designing different utility functions for different types of traffic,the routing scheme quantifies the degree of satisfaction of the path to traffic QoS as a utility value,and performs path calculations based on the utility value,thereby guaranteeing the QoS requirements of different types of traffic flows.At the same time,the routing scheme can predict the link traffic based on the LSTM neural network,and can judge and prevent the link congestion risk based on the result of the traffic prediction.Second,a QoS routing scheme based on reinforcement learning is designed.The routing is based on the DDPG deep reinforcement learning algorithm,which can learn spontaneously according to the customized QoS optimization strategy and the network status obtained from the SDN network,and generate routing strategies based on the learned knowledge,thereby maximizing the comprehensive network utility.Third,QoS routing based on utility value and traffic prediction and QoS routing based on reinforcement learning are implemented in ONOS controller,and the function and performance of QoS routing are tested by building an SDN network simulation platform based on Mininet.The test results show that QoS routing based on utility value and traffic prediction can effectively control network congestion and better guarantee the QoS requirements of different services;QoS routing based on reinforcement learning has good convergence and can effectively reduce the end-to-end delay and packet loss rate of the traffic flow.