Research On SDN Load Balancing Based On Ant Colony Optimization Algorithm

With the continuous development of cloud computing and big data,the scale of the data center has been further expanded.The services provided by the data center have begun to be diversified,resulting in the rapid growth of network traffic.Therefore,the management of the underlying network of the data center is more difficult.Software-defined network(SDN)is a new network paradigm that has emerged in recent years.Due to its advantages of global view and network programmability,it is increasingly deployed in data centers.Therefore,the load balancing of SDN data centers has gradually become a research hotspot in the computer field.The paper focuses on SDN data center load balancing research.Through the research of server load balancing and link load balancing,a DLBS algorithm is proposed.The DLBS algorithm mainly includes two stages.The first stage is server load balancing.Firstly,the server load information is collected by sFlow deployed in the SDN controller,and then the server processing power and the current load are comprehensively considered,the weight of the server is dynamically set,and finally the data flow select the appropriate server for processing based on the weight.The second phase is link load balancing.Aiming at the problem that the traditional ECMP load balancing solution is easy to cause the collision of the elephant stream,a dynamic rerouting strategy is proposed.The data link and the switch status information obtained by the SDN controller are used to evaluate the path,and whether to perform the rerouting operation according to the evaluation result is selected.In the process of rerouting,the paper adopts the ant colony optimization algorithm.In view of the shortcomings of the basic ant colony algorithm,the global search ability is weak and the convergence speed is slow.Three improvements are proposed: fusion of chaotic selection strategy,combination of pheromone local update and global update,segmentation adjustment of pheromone volatilization factor.In order to verify the availability and effectiveness of the DLBS algorithm,Iperf generates different traffic injections into the custom network topology for experimental testing.The experimental results show that compared with the polling algorithm and the random algorithm,the DLBS algorithm has larger system throughput,shorter response time and more balanced CPU utilization.In the link load balancing,the DLBS algorithm is compared with the shortest path,the average transmission delay and average UDP packet loss rate of the system is lower,and the average stream throughput is larger.The experimental results verify the effectiveness of the DLBS algorithm.