Research On Load-balancing Strategy Based On The Software Defined Network

The number of visits in the network has been growing aggressively with the rapid development of the Internet.The traditional network architecture with IP as the core can not meet the needs of people due to the problems of its scalability,security and managing ability.Software Defined Networking(SDN)is one of the key technologies of the future network,which abstracts the control functions from the switch,separates the forwarding and control,and has the features of centralized control and programmablity.This thesis mainly studies a key problem in SDN architecture-load balancing,summarizes the advantages and disadvantages of existing load balancing algorithms,and puts forward load balancing strategy based on forwarding plane and controller plane respectively.Firstly,a load balancing routing strategy is proposed based on utility maximize by taking advantage of centralized control of OpenFlow network controller.The number of paths that exist between the source node and the destination node are determined first in the strategy.Different weights are given to the relevant parameters of Quality of Service(QoS)according to different attributes of the data stream.And impact factor value of network quality of service such as link utilization,delay,and packet loss rate is acquiring by gathering the information from the switch.And then using the above values to calculate the utility and load of each path in the path set.Finally,the preferred path for the data flow is figured out through integrating overall considerations.This strategy can satisfy the high quality of service to the greatest extend while ensuring the load balancing of network.Simulation experiment is designed and the experiment results prove that the proposed load balancing strategy in this thesis can figure out the optimal path for the data flow,alleviate the congestion of networks effectively,and improve the overall performance of the network.Secondly,excessive request of the underlying switch results in controller overload in the distributed controller network architecture.To solve the above problem,a dynamic load balancing strategy among the controllers is proposed.The controller load is calculated by real-time statistics in the strategy.When the controller is overloaded,the migration cost of the switch in the overload controller is calculated and migrated to other suitable controllers according to the migration cost,thus reducing the load of the overload controller.And the load on the overload controller is reduced to the average load level of the network controller in order to avoid new load imbalances due to other controllers receiving too much load.Simulation experiment is designed and the experiment results prove that the proposed strategy can reduce the load of the overload controller effectively,so that the load among the controllers in the network is more balanced.