Research On SDN Load Balancing Strategy Based On Improved Artificial Bee Colony Algorithm

With the rapid growth of the network scale,the data center network has also expanded significantly.Due to the problems of poor controllability and the closedness of network equipment in the traditional data center network,researchers have considered applying Software Defined Network(SDN)technology to Data center network.How to circumvent some of the problems caused by the centralized control method of SDN and give full play to the advantages of SDN in traffic scheduling has become the key to research.This article studies the load balancing strategy applicable to the current data center network from the two entry points of link and switch.The main contents are as follows:1.A load balancing algorithm based on improved artificial bee colony algorithm(IALB)is proposed to solve the problems of data center network instability and link congestion caused by traditional methods of scheduling elephant flow.The IALB algorithm is based on the design of artificial bee colony algorithm and defines the concept of link stability.It redefines the fitness function f(xi)of the artificial bee colony algorithm based on the real-time remaining bandwidth of the link,the number of large flows and the link stability.This improvement of the artificial bee colony algorithm results in the approximate optimal scheduling plan of the elephant flow.The simulation experiment results show that the IALB algorithm proposed in this paper is better than the ECMP,Hedera and PureSDN algorithms in terms of network throughput and link utilization performance indicators,and has a significant improvement in the two communication modes of Random and Staggered.2.Aiming at the problem that the existing algorithms cannot fully utilize the flow table resources and controller resources in the SDN network,a load balancing algorithm based on flow table resources and controller resources(FCLB)is proposed.The FCLB algorithm uses the invalid time of the flow table entry to represent the resource consumption of the flow table,uses the number of packet-in messages to represent the resource consumption of the controller,and defines the concept of resource preference factor.Based on the resource preference factor,the relationship between idle_timeout and resource consumption is established,by dynamically setting the value of idle_timeout to achieve balanced utilization of flow table resources and controller resources.The simulation experiment results show that the FCLB algorithm is better than the original solution in terms of flow table resource utilization and data flow hit rate.It can be seen that the FCLB algorithm can make better use of flow table resources and controller resources,thereby improving network throughput.