摘要
给出一种基于软件定义网络(software defined network,SDN)控制器的负载均衡策略。选用分布式控制器,将网络中任意两节点的通信近似为Markov过程,通过多步转移概率计算网络节点间各链路权重。测量交换机和控制器负载,当控制器超载后,根据链路权重,选择需要迁移的交换机进行负载均衡操作,最后在数据中心Fat-Tree拓扑上进行网络仿真。仿真结果表明,与最短路径优先算法相比,所给策略的平均流量和吞吐量均有明显提升。
A load balancing strategy based on SDN (software defined network ) controller is proposed to solve the network node load imbalance problem. By using the distributed controller, the communication approximation of any two nodes in the network is regarded as the Markov process. The link weights between network nodes are calculated by multi-step transition probabilities. Measuring the load of the switches and the co ntrol le rs, when the controller is overloaded, the load balancing operation is selected according to the l in k weight. F in a l ly , the strategy is running on data centre Fat-Tree topology. Simulation results show that the proposed load balancing strategy can improve the average traffic and throughput of the network nodes compared with the shortest path firstalgor ithm .
出处
《西安邮电大学学报》
2017年第4期38-42,共5页
Journal of Xi’an University of Posts and Telecommunications
基金
陕西省教育厅科学研究计划资助项目(07JK377)
关键词
负载均衡
软件定义网络
马尔可夫过程
load balancing, software defined network , Markov process