基于蚁群算法的SDN数据中心网络大象流调度研究  

Flow scheduling of elephant flows in SDN data center network based on ant colony algorithm

作　　者：李宏慧[1] 杨光 路海亮 付学良[1] 申志军[1] Li Honghui;Yang Guang;Lu Hailiang;Fu Xueliang;Shen Zhijun(College of Computer&Information Engineering,Inner Mongolia Agricultural University,Hohhot 010010,China)

机构地区：[1]内蒙古农业大学计算机与信息工程学院

出　　处：《计算机应用研究》2019年第12期3837-3841,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61363016);国家教育部留学回国人员科研启动基金项目([2014]1685);内蒙古自治区自然科学基金资助项目(2015MS0605,2015MS0626)

摘　　要：针对传统方法调度大象流时容易造成数据中心网络拥塞和负载不均衡等问题,提出一种基于蚁群算法的SDN(software defined network)数据中心网络流量调度算法ACO-SDN。对大象流调度问题建立整型线性规划ILP(integral linear programing)模型,优化目标为最小化最大链路利用率。通过重定义蚁群算法的参数和操作求解ILP模型,得到大象流重路由的最优路径。实验结果表明,与ECMP(equal-cost multi-path routing)和GFF(global first fit)流量调度算法相比,ACO-SDN算法降低了网络最大链路利用率,有效地提高了网络对分带宽。The traditional methods may cause network congestion and load imbalance when scheduling the elephant flows in data center networks. To deal with these issuses,this paper proposed scheme ACO-SDN,which was based on ACO( ant colony)algorithm in SDN data center networks. It built an ILP model according to the flow scheduling problem,with the object of minimize the maximum link utilization. Then,it employed ant colony algorithm to solve the ILP model through redefining its parameters and operations. It obtained the optimal route of rerouting elephant flows. The experimental results show that,in comparisons with ECMP( equal-cost multi-path routing) and GFF( global first fit) algorithms,ACO-SDN algorithm lowers down the maximum link utilization rate,and effectively improves the network bisection bandwidth of networks.

关 键 词：数据中心网络 软件定义网络 大象流调度 蚁群算法 

分 类 号：TP393.07[自动化与计算机技术—计算机应用技术]