Research On Load Balancing Strategy Based On SDN Controller And Traffic Scheduling

In recent years,with the development of domestic e-commerce and live broadcasting industry,the scale of network traffic has also increased rapidly,which has put great pressure on the traffic transmission of the network.In this context,the traditional IP architecture network has been unable to adapt to the current network environment due to the shortcomings of bloated architecture,cumbersome configuration,and difficult maintenance.Software defined network(SDN)architecture,thanks to its centralized control logic and strong network programmability,is not only more able to adapt to complex network environments,but also provides unique conditions for the development of network technology.SDN has become a hot topic and focus of academic and industrial research,and it is also being deployed on a large scale in the real world,especially in data center networks.However,due to the lack of flexible and effective load balancing solutions in the current SDN architecture,the performance of the network is limited,so it is urgent to improve these problems.For the control plane,the introduction of a multi-controller architecture improves the scalability of the control plane,but the issue of load imbalance between controllers limits the resource utilization of the control plane.For the data plane,the data center network traffic is complex and the topology is special,and the traditional link load balancing algorithm cannot balance the link load in this environment,resulting in network congestion and inability to make full use of network resources.According to the above analysis,the research work in this paper can be summarized into the following two aspects:(1)For control planes with multiple controllers,the static connection between controllers and switches may cause the load of some controllers to be large or even overloaded,causing load imbalance between multiple controllers,and this thesis gives a multi-controller dynamic load balancing mechanism(MDLB).When the controller load variance exceeds the threshold,the super controller trades off migration efficiency and migration cost to develop an appropriate switch migration strategy to achieve load balancing of the control plane.Experimental results show that the MDLB mechanism can effectively improve the load balancing degree and throughput of the control plane,and reduce the average migration time of the switch.(2)Aiming at the link load imbalance problem caused by the inability of the traditional link load balancing algorithm ECMP to adapt to the data center network,this thesis gives a particle swarm optimization algorithm based dynamic multipath load balancing mechanism(PDMLB).When the load balancing factor exceeds the threshold,the controller schedules the large-flow detected and marked by the host side to the appropriate path according to the global topology information collected by the Open Flow protocol and combined with the particle swarm optimization algorithm,so as to avoid the collision of the large-flow and achieve load balancing.Simulation results show that compared with ECMP algorithm and SA algorithm,this mechanism can effectively improve the throughput of data plane and network bandwidth utilization,and reduce network latency.