Tianshi Wang, Hongwei Kan, Qibo Sun, Shan Xiao, Shangguang Wang
{"title":"Congestion Detection and Link Control via Feedback in RDMA Transmission","authors":"Tianshi Wang, Hongwei Kan, Qibo Sun, Shan Xiao, Shangguang Wang","doi":"10.1109/ICSS55994.2022.00010","DOIUrl":null,"url":null,"abstract":"Researchers and practitioners are exploiting Remote Direct Memory Access (RDMA) technology to improve the efficiency of distributed machine learning and meet the demands of data-center applications. RDMA requires lossless network link to fully unleash its power. RDMA Over Converged Ethernet (RoCE) v2 focuses on congestion control, but fails to achieve efficient packet loss recovery; Improved RoCE NIC (IRN) addresses this issue based on RoCEv2, but does not use the Priority-based Flow Control (PFC) to maintain the advantage of RoCEv2 in detecting congestion. This paper proposes a method of congestion detection and link control via feedback in RDMA transmission, namely Feedback Data Flow Control (FDFC), that does not rely on PFC. FDFC detects and controls the link condition in real time to achieve the goals of precise detection, congestion control, and efficient packet loss recovering.","PeriodicalId":327964,"journal":{"name":"2022 International Conference on Service Science (ICSS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 International Conference on Service Science (ICSS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICSS55994.2022.00010","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
Abstract
Researchers and practitioners are exploiting Remote Direct Memory Access (RDMA) technology to improve the efficiency of distributed machine learning and meet the demands of data-center applications. RDMA requires lossless network link to fully unleash its power. RDMA Over Converged Ethernet (RoCE) v2 focuses on congestion control, but fails to achieve efficient packet loss recovery; Improved RoCE NIC (IRN) addresses this issue based on RoCEv2, but does not use the Priority-based Flow Control (PFC) to maintain the advantage of RoCEv2 in detecting congestion. This paper proposes a method of congestion detection and link control via feedback in RDMA transmission, namely Feedback Data Flow Control (FDFC), that does not rely on PFC. FDFC detects and controls the link condition in real time to achieve the goals of precise detection, congestion control, and efficient packet loss recovering.
研究人员和实践者正在利用远程直接内存访问(RDMA)技术来提高分布式机器学习的效率，满足数据中心应用的需求。RDMA需要网络链路无损，才能充分发挥其威力。RDMA Over Converged Ethernet (RoCE) v2侧重于拥塞控制，但无法实现高效的丢包恢复;改进的RoCE NIC (IRN)在RoCEv2的基础上解决了这个问题，但没有使用基于优先级的流量控制(PFC)来保持RoCEv2在检测拥塞方面的优势。本文提出了一种在RDMA传输中通过反馈进行拥塞检测和链路控制的方法，即反馈数据流控制(feedback Data Flow control, FDFC)，它不依赖于pfc，实时检测和控制链路状态，以达到精确检测、拥塞控制和高效丢包恢复的目的。
京公网安备 11010802042870号
京ICP备2023020795号-1
分享
求助内容：
应助结果提醒方式：
扫码关注我们
