基于SDN的两级大象流负载均衡策略

Load Balancing Strategy of a Two-stage Elephant Flow Based on SDN

以SDN网络为基础,提出一种解决数据中心网络拥塞的大象流负载均衡策略,完善了负载均衡整体框架.第一阶段使用sFlow收集网络状态,通过降低大象流误检率,采用突出大象流漏检率并设置阈值以判定可疑大象流;第二阶段应用基于流持续时间判定真正大象流,剔除已变质大象流,并将两阶段的大象流检测方法嵌入整体负载均衡架构中;最后在ubuntu系统环境下利用Mininet搭建胖树网络拓扑,以Ryu为控制器进行仿真模拟测试,通过控制流量负载验证大象流负载均衡策略.试验结果表明,提出的大象流负载均衡策略可提高数据中心网络中大象流与老鼠流的检测效率和网络链路利用率,并可减小网络传输时延.

Based on the SDN network,an elephant flow load balancing strategy for solving the data center network congestion is proposed,and the overall framework of load balancing is improved.The first phase uses sFlow to collect network status,and reduces the false detection rate of elephant flow.The misdetection rate of elephant flow is highlighted and a threshold is set to determine suspicious elephant flow.The second stage is based on the flow duration to determine the true elephant flow and eliminate the metamorphic elephant flow,then a two-stage elephant flow detection method is embed into the overall load balancing.Finally,in the ubuntu system environment,Mininet is used to build the fat tree network topology,and Ryu is used as the controller for the simulation test.The elephant flow load balancing strategy is verified by controlling the load flow.The experimental results show that the proposed elephant flow load balancing strategy can improve the detection efficiency both of the elephant and mouse stream in the data center network and the network link utilization.At the same time,the network transmission delay can be reduced as well.