一种软件定义网络中交换机动态迁移算法  

Dynamic switches migration algorithm in software defined networks

作　　者：陈飞宇[1] 汪斌强[1] 孟飞[1] 王雨薇[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,郑州450002

出　　处：《计算机应用研究》2016年第5期1446-1449,1480,共5页Application Research of Computers

基　　金：国家"973"计划资助项目(2012CB315901;2013CB329104);国家自然科学基金资助项目(61372121);国家"863"计划资助项目(2013AA013505)

摘　　要：针对目前软件定义网络中的多控制器负载失衡问题,提出一种交换机的动态迁移算法(dynamic switches migration algorithm,DSMA),将交换机与控制器的部署关系建模为0-1规划问题,通过使用免疫粒子群算法保证控制器负载均衡的同时兼顾了控制器和交换机之间的传输时延。仿真实验表明,与现有经典的就近迁移算法和利用率最低迁移算法相比,DSMA实现了较好的控制器负载均衡,减少了控制器PACKET_IN消息的响应时间,提高了系统反应速度,加权后的综合评价平均提升了25.3%。Aiming at the load imbalance of multiple controllers in software defined networks,this paper proposed a dynamic switches migration algorithm（ DSMA）. By modeling the controller-switch mapping as 0-1 programming problem,this algorithm used the immune particle swarm algorithm to achieve the load balance of the controllers and ensured the transmission delay between controllers and switches at the same time. Simulation results show that,compared with the existed typical algorithms such as nearest migration algorithm and lowest utilization migration algorithm,DSMA achieves good load balancing of controllers,reduces the response time of the PACKET_IN messages,improves the system response speed and increases the weighted comprehensive evaluation meanly to 25. 3%.

关 键 词：软件定义网络 控制器 负载均衡 免疫粒子群 

分 类 号：TP393.06[自动化与计算机技术—计算机应用技术] TP301.6[自动化与计算机技术—计算机科学与技术]