首页 | 本学科首页   官方微博 | 高级检索  
     

数据中心网络中的流量均衡
引用本文:李兆耕,毕军. 数据中心网络中的流量均衡[J]. 软件学报, 2016, 27(S2): 243-253
作者姓名:李兆耕  毕军
作者单位:清华大学 网络科学与网络空间研究院, 北京 100084;清华大学 计算机科学与技术系, 北京 100084;清华信息科学与技术国家实验室(筹)(清华大学), 北京 100084,清华大学 网络科学与网络空间研究院, 北京 100084;清华大学 计算机科学与技术系, 北京 100084;清华信息科学与技术国家实验室(筹)(清华大学), 北京 100084
基金项目:国家自然科学基金(61472213)
摘    要:现代数据中心网络在任意两个主机之间都存在很多可选路径.如何在多个可选路径之间实现流量均衡,是数据中心网络中的重要研究课题.针对这一问题,已有研究者提出了很多解决方案.对多级Clos架构下的数据中心网络中的流量均衡问题做了深入的分析与总结.首先分析了数据中心网络的特点,然后定义流量均衡问题为最小化等价链路的最大潜在丢包率.之后总结了各种丢包产生的原因,并讨论了影响流量均衡设计方案的两个主要挑战:分组乱序与突发拥塞.在此基础上,把现有解决方案分为主动调度、切片散射、探测与调整及其他这4个大类,并对各解决方案逐一进行介绍,说明各自的优缺点.最后还对不同方案作了对比,指出了未来可能的研究方向.

关 键 词:数据中心网络  流量均衡  ECMP  潜在丢包  突发拥塞
收稿时间:2016-06-05
修稿时间:2016-10-18

Traffic Balancing in Data Center Networks
LI Zhao-Geng and BI Jun. Traffic Balancing in Data Center Networks[J]. Journal of Software, 2016, 27(S2): 243-253
Authors:LI Zhao-Geng and BI Jun
Affiliation:Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China;Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China;Tsinghua National Laboratory for Information Science and Technology(TNList)(Tsinghua University), Beijing 100084, China and Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China;Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China;Tsinghua National Laboratory for Information Science and Technology(TNList)(Tsinghua University), Beijing 100084, China
Abstract:There are many optional paths between any server pair in modern data center networks. Traffic balancing among the paths is an important issue. Many solutions have been proposed to cope with this problem. This paper analyzes traffic balancing in data center networks with multi-tier Clos-based topology. The feature of data center networks is introduced and the traffic balancing problem is defined as minimizing the maximum potential packet loss among different equivalent links. Then the reasons of potential packet loss are summarized in data center networks and two main challenges are discussed related to traffic balancing:packet out-of-order and burst congestion. Ten existing solutions are classified into four categories:active scheduling, slice spraying, probing and adjusting, and others. The advantages of these solutions are described one by one in details, as well as their disadvantages. A comparison is made among these solutions and possible research directions are pointed out on traffic balancing in data center networks.
Keywords:data center network  traffic balancing  ECMP  potential packet loss  burst congestion
点击此处可从《软件学报》浏览原始摘要信息
点击此处可从《软件学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号