数据中心网络TCP Incast问题研究

发布时间：2018-06-16 05:32

本文选题：数据中心网络 + TCP　；参考：《计算机科学与探索》2017年09期

【摘要】：目前传统TCP协议不适用于数据中心的工作模式,因此当数据中心中出现常见的多对一流量模式时会产生TCP Incast问题,造成应用层可见吞吐量崩溃。结合数据中心特点,提出全面的解决方案是解决TCP Incast问题的研究目标。围绕TCP Incast问题,深层次剖析了该问题发生的根源,简要概述了该问题面临的挑战,介绍了基于该问题所构建的数学模型;从链路层、传输层和应用层角度分析并总结了近十年具有代表性的解决方案,从有效性、可部署性等不同角度对所列举方案进行了全面对比,发现当前方案大都基于某个具体方面缓解该问题,均存在缺陷;最后提出了可行的解决该问题的研究方向,将关注点聚焦于SDN结合机器学习以及传输新协议。
[Abstract]:At present, the traditional TCP protocol is not suitable for the working mode of the data center, so when the common many-to-one traffic mode appears in the data center, the TCP cast problem will occur, resulting in the collapse of the application layer visible throughput. According to the characteristics of data center, a comprehensive solution is the research goal of solving TCP cast problem. Around the TCP cast problem, this paper deeply analyzes the root cause of the problem, summarizes the challenges it faces, and introduces the mathematical model based on the problem. This paper analyzes and summarizes the representative solutions in the last ten years from the point of view of transport layer and application layer, and makes a comprehensive comparison of the listed solutions from different angles, such as effectiveness, deployable ability, etc. It is found that most of the current schemes are based on a specific aspect to alleviate the problem, and there are defects. Finally, a feasible research direction to solve this problem is proposed, focusing on SDN combined with machine learning and new transport protocols.
【作者单位】：清华大学计算机科学与技术系;
【分类号】：TP308;TP393.0

【相似文献】