| The proposal of Software Defined Network(SDN)satisfies the rapidly increasing demand for internet equipments and data scale.With the explosive growth of the network scale and the continuous iteration of applications,the SDN control plane is carrying heavier and heavier loads.Load imbalance problems caused by controller overloads occur from time to time,and the bottleneck problems in the control plane are becoming more and more serious,which seriously affects SDN network performance.Therefore,how to allocate controller resources to maximize network performance has become a key issue in the study of controllers.This article proposes two solutions to the problem of SDN multi-controller load imbalance:In the case of single controller overload,based on the principle of minimizing load migration delay and network overhead,a reverse auction SDN multi-controller load balancing algorithm is proposed,SMLBRA;When multiple controllers in the network are overloaded at the same time,the SMLBRA algorithm will reduce the switch migration efficiency and increase the network overhead cost.Therefore,a SDN multi-controller load balancing algorithm based on double auction,SMLBDA,is proposed.(1)Design and implement SDN multi-controller load balancing algorithm based on reverse auction(SMLBRA).First,design the model of the SMLBRA algorithm,and combine the reverse auction theory with the SDN multi-controller architecture to establish a "one-to-many" reverse auction model.In the model,the overloaded single controller is regarded as the buyer,and the lightly loaded multiple controllers as a seller,the auctioneer is composed of three modules: load monitoring module,load integration module and load decision module.Secondly,the SMLBRA algorithm is analyzed and designed in detail.In the load decision module,the auctioneer selects the appropriate seller as the target controller based on the seller’s bid,and combines the simulated annealing algorithm to minimize the average delay and network overhead as the optimization goal.Obtain the remapping of the controller and the switch,so that the controller resources can be reasonably allocated.Finally,an experimental simulation is carried out to compare and analyze the SMLBRA algorithm with STATIC,ECSM,and SMDM algorithms in terms of controller resource utilization,system response delay and load migration completion time.(2)Design and implement SDN multi-controller load balancing algorithm based on double auction(SMLBDA).First,according to the "many-to-many" controller overload problem,the SDN multi-controller model based on double auction is constructed,which is composed of buyer-overload controller,seller-light-load controller and auctioneer.Then the SMLBDA algorithm is divided into two stages: the multi-controller resource allocation stage and the switch migration stage.Multi-controller resource allocation stage: The load monitoring module in the model periodically monitors the load of the controller,sorts and bids according to its overload degree.The higher the overload degree,the higher the bid price.Similarly,the light-load controller according to its The remaining resources are offered for sale.Combining genetic algorithm with the goal of maximizing network benefits,the overload controller is matched with the target controller.Switch migration stage: Complete the switch migration process by minimizing the impact of switch migration on network performance,so as to achieve control plane load balancing.Finally,through the experimental analysis of the algorithm,the SMLBDA algorithm is compared with the PASMM and DALB algorithms,and the effectiveness of the algorithm is verified in terms of controller resource utilization and load balancing time.Mininet simulation experiments show that the SMLBRA algorithm proposed in this paper has improved performance in reducing controller resource utilization and response delay,and the SMLBDA algorithm has improved performance in reducing load balancing time. |
