基于时延的趋势感知的无损数据中心网络拥塞控制  

Delay-Based Trend-Aware Congestion Control in Lossless Datacenter Network

作　　者：黄凯欣 HUANG Kaixin(Shanghai Yunbao Chuangxin intelligent Technology Co.,Ltd.,Shanghai 201306)

机构地区：[1]上海云豹创芯智能科技有限公司,上海201306

出　　处：《中国科技纵横》2024年第8期52-57,133,共7页China Science & Technology Overview

摘　　要：在当前的高速数据中心网络中,拥塞控制对于保证持续的高性能十分重要。在过去十年里,研究人员和开发人员已经探索了多种拥塞信号,比如ECN(显式拥塞通知)、RTT(往返时间)和INT(内部网络遥测)。然而,现有的大多数拥塞控制算法要么因为信号含糊而导致拥塞检测不精确,要么因为过于激进的降速而导致过多的带宽损失。本文提出了一种名为DELTA的新型拥塞控制机制,这是一种为无损数据中心网络设计的基于延迟的趋势感知方法。DELTA通过利用RTT的变化识别拥塞趋势,并相应地调整发送速率。通过分析拥塞趋势,发送端将发送速率调整至合理水平应对拥塞,同时仍然保持高带宽利用率消除拥塞。基于此,使用NS-3仿真平台对DELTA进行广泛评估。实验结果表明,DELTA在流量完成时间(FCT)和收敛速度方面均优于拥塞控制算法。In current high-speed datacenter networks,congestion control is crucial for ensuring consistent high performance.Over the past decade,researchers and developers have explored several congestion signals such as ECN,RTT,and INT.However,most of the existing congestion control algorithms suffer from either imprecise congestion detection due to ambiguous signals or excessive bandwidth loss due to aggressive rate decrease.This paper proposes a novel congestion control mechanism called DELTA,which is a delay-based trend-aware approach designed for lossless datacenter networks.DELTA leverages the change in RTT to learn the congestion trend and adjusts the sending rate accordingly.By analyzing the congestion trend,the sender reacts by adjusting the sending rate to a reasonable level,while still maintaining high bandwidth utilization to dismiss congestion.We evaluate DELTA extensively in NS-3 simulations and the experimental results demonstrate that DELTA outperforms the compared congestion control algorithms in both FCT and convergence speed.

关 键 词：拥塞控制 无损网络 网络仿真 流量工程 数据中心网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]