Research On Dynamic Traffic Scheduling Technology Of Datacenter Network Based On SDN

With the rise of cloud services and the continued virtualization of data centers,datacenter networks are rapidly evolving.Traditional datacenter network technologies are no longer able to meet the increasing workload and scale requirements.The datacenter network based on software-defined network which is dynamically programmed has gradually become the development trend and the mainstream of the industry.In the emerging data center online business,due to the widespread use of distributed technologies,the proportion of east-west traffic is expanding.How to effectively manage the traffic transmission between servers and control the network traffic in data center is an urgent problem to be solved.SDN-based datacenter network traffic dynamic scheduling technology is deeply researched,and current datacenter network topology structure and traffic characteristics are also analyzed.SDN-based single-controller hierarchical dynamic traffic scheduling system is implemented.Multi-controller load balancing system is realized by extending single controller.The main work includes:(1)Aiming at the problem of low bandwidth resource utilization and high traffic transmission delay in datacenter network,a SDN-based hierarchical dynamic traffic scheduling algorithm is proposed.Modify and extend the source code of network controller to periodically collect link information such as bandwidth and delay.Asynchronously detect congestion links and reschedule network traffic.Monitor forwarding ports to classify network traffic.The k-shortest path algorithm is used to implement multi-path calculation,which reduces the range of link evaluation.Apply multi-path evaluation to the traffic with different characteristics to achieve optimal path forwarding.The hierarchical dynamic traffic scheduling algorithm is applied to RYU controller to realize single controller traffic scheduling system.At the same time,GLB algorithm is also implemented in the system for comparison experiments.The experimental results show that hierarchical dynamic traffic scheduling algorithm can reduce the transmission delay of 58.09% and improve the network bandwidth utilization by 7.17% compared with GLB algorithm.(2)For a single controller that unable to meet the network expansion and multiple controllers cannot achieve load balance,a layered multi-controller load balance system is implemented,and the upper layer uses the load balance host to centrally control the lower layer controllers.In the multi-controller system,each controller is responsible for an autonomous domain,and a single-controller hierarchical dynamic traffic scheduling system is applied in the autonomous domain.The migration of compute task is implemented through the upper-layer load balance host among multiple autonomous domains.This enables load balance of multiple controllers.At the same time,the load balance host is responsible for the construction of global network topology view.Topology sharing and cross-domain communication are also realized.Based on the layered network architecture design,a multi-controller load balance prototype system is implemented.Experiments are carried out in the simulation environment.The experimental results show that the system can achieve coordination and load balance between multiple controllers while enhancing scalability.