| With the rapid development of next-generation network technology,the traffic of data center networks has increased dramatically and become more complex,and traditional networks have exposed the defects of insufficient link utilization and unpredictable network status,making it difficult to load balance and optimize the routing of data center networks in a flexible manner.SDN decouples the control plane from the data plane and can realize unified control,configuration and management of the network through a global view,which provides a new idea to improve the performance optimization of data center network in many aspects.Therefore,the focus of this thesis is to design an efficient load balancing and routing optimization strategy based on SDN,and the following two parts are studied.First,for the problems of traffic conflict,hash collision and inability to fully utilize link resources in ECMP load balancing strategy,a load balancing system for data center network based on SDN is designed,which can sense the network state in real time and dynamically divide the data flow into small and large flows.Based on the different transmission characteristics of large and small streams,a load balancing strategy(LBSPSO)based on particle swarm optimization algorithm is then proposed,and the concepts of load distribution degree and path congestion degree are introduced to achieve uniform distribution of large streams by improving the design of particle swarm algorithm.The improved KSP algorithm is used to ensure the fast transmission of small flows.In this thesis,a simulation platform based on ONOS and Mininet is built.The results show that the proposed strategy can effectively improve the load balancing effect,increase the link bandwidth utilization,and reduce the packet loss rate of small streams compared with ECMP and Hedera strategies.Second,for the multiple QoS requirements of video streaming services,a routing optimization strategy for video streaming QoS guarantee is proposed,which obtains the available bandwidth,experiments,packet loss rate and jitter indexes in the network through the SDN controller,proposes a routing optimization algorithm based on reinforcement learning,and uses an intelligent body to autonomously explore and gain experience in the network environment to avoid network congestion as much as possible,so as to select the best The results show that this strategy is able to achieve the optimal transmission path in different network environments.The results show that this strategy can effectively reduce the packet loss rate and jitter and improve the network throughput under different traffic load pressure.There is also a certain performance improvement in terms of delay,which guarantees the QoS requirement of video stream transmission and verifies the effectiveness and feasibility of the strategy. |
