基于SDN的云架构网络高确定性流量控制方法  

High Deterministic Traffic Control Method for Cloud Architecture Network Based on SDN

作　　者：高平[1] 张帆 张东 翟飞龙 GAO Ping;ZHANG Fan;ZHANG Dong;ZHAI Feilong(China National Aeronautical Radio Electronics Research Institute,Shanghai 200233,China;Department of Computer Science and Technology,Xi’an Jiaotong University,Xi’an 710049,China;State Key Laboratory of Geographic Information Engineering,Xi’an 710054,China;The 32nd Research Institute of China Electronics Technology Group Corporation,Shanghai 201808,China)

机构地区：[1]中国航空无线电电子研究所,上海200233 [2]西安交通大学计算机科学与技术系,西安710049 [3]地理信息工程国家重点实验室,西安710054 [4]中国电子科技集团公司第三十二研究所,上海201808

出　　处：《计算机工程》2018年第12期74-78,84,共6页Computer Engineering

基　　金：中国航空无线电电子研究所项目(PM-12210-2016-001)

摘　　要：针对目前云架构网络不能较好地应对云网络中数据传输的多样性以及细粒度和确定性流量控制需求等问题,提出一种基于软件定义网络(SDN)的高确定性流量控制方法。将SDN集中优化控制、全网感知能力的优势与交换机上的多队列调度机制相结合,在获取全网拓扑和流量信息的基础上规划业务流最优路径,为交换机分配合理的队列调度参数,并利用DCTCP减小网络拥塞,保证各类业务的带宽和时延,同时运用全可编程交换机实现业务流的细粒度管理,配合个性化的流表下发机制,动态调整业务传输带宽和业务流等级。实验结果表明,该方法关键业务流的平均排队时延小于2 ms,排队时延抖动小于10 ms,丢包率小于10^(-6),可实现对云网络中的业务流集中化、可实时化和精细化管理。Aiming at the problem that the current cloud architecture network can not deal with the diversity of data transmission in cloud network and the requirements of fine-grained and deterministic traffic control,a high-deterministic traffic control method based on Software Defined Network(SDN)is proposed.It combines the advantages of centralized optimization control of SDN and the awareness of the whole network with the multi-queue scheduling mechanism on the switch,and plans the optimal path of the service flow based on the topology and traffic information of the whole network,and allocates a reasonable queue scheduling parameters for the switch,and uses DCTCP to reduce network congestion and ensures bandwidth and delay of various services.At the same time,the full programmable switch is used to implement the fine-grained management of the service flow,and the personalized flow table delivery mechanism is used to dynamically adjust the service transmission bandwidth and service flow level.Experiment result shows that the critical business flow of this method achieves the average queueing delay less than 2 ms,the queueing delay jitter is less than 10 ms,and the packet loss rate is less than 10-6,it can realizes the centralized,real-time and fine management of the traffic flow in the cloud network.

关 键 词：云架构网络 流量控制 软件定义网络 多队列调度 服务质量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]