数据中心网络拥塞控制机制的设计与实现  

Design and Implementation of Congestion Control Mechanism in Data Center Network

作　　者：翟烈烈 汪学舜 戴锦友[2] ZHAI Lielie;WANG Xueshun;DAI Jinyou(Wuhan Research Institute of Posts and Telecommunications Wuhan,430074,China;Fiberhome Telecommunication Technologies Co.,Ltd.,Wuhan,430074,China)

机构地区：[1]武汉邮电科学研究院,武汉430074 [2]烽火通信科技股份有限公司,武汉430074

出　　处：《网络新媒体技术》2024年第2期61-67,共7页Network New Media Technology

基　　金：科技部重大研发专项:多模态网络控制调度系统技术(编号:2022YFB2901200)。

摘　　要：拥塞控制技术是数据中心网络中实现超低延迟、高带宽和网络稳定的关键。现有的拥塞控制技术难以同时兼顾低延迟、高带宽和网络稳定。因此,提出一种基于网络感知的拥塞控制机制,利用网络带内遥测技术获取精确的链路负载信息,精确控制流量。通过解决拥塞期间的延迟信息和对信息的过度反应等问题,实现快速收敛以利用空闲带宽,避免拥塞。测试结果表明在小型数据中心网络中,采用本文设计的拥塞控制机制与dcqcn算法进行比较,当流量较小时,在前99%流量的情况下流放缓比的值前者仅为后者的10%左右。Congestion control technology is the key to achieve ultra-low latency,high bandwidth and network stability in data center networks.It is difficult for the existing congestion control technology to combine low latency,high bandwidth and network stability at the same time.In this paper,a congestion control mechanism based on network awareness is proposed,which uses network in-band telemetry to obtain accurate link load information and control traffic accurately.By solving problems such as delayed information and overreaction to information during congestion,fast convergence is realized to utilize idle bandwidth and avoid congestion.The test results show that in small data center networks,when the congestion control mechanism designed in this paper is compared with the dcqcn algorithm,the flow slowdown ratio of the former is only about 10%of the latter in the case of the first 99%traffic.

关 键 词：网络感知 拥塞控制 数据中心网络 超低延迟 链路负载 

分 类 号：TP308[自动化与计算机技术—计算机系统结构]