基于分布式优化的数据中心网络混流调度机制

张彤, 任丰原, 舒然

清华大学学报（自然科学版） ›› 2021, Vol. 61 ›› Issue (6) : 618-625.

PDF(1541 KB)
横山亮次奖  |  百年刊庆  |  中文  |  English
PDF(1541 KB)
清华大学学报（自然科学版） ›› 2021, Vol. 61 ›› Issue (6) : 618-625. DOI: 10.16511/j.cnki.qhdxxb.2020.21.018
计算机科学与技术

基于分布式优化的数据中心网络混流调度机制

  • 张彤1,2, 任丰原3, 舒然4
作者信息 +

Distributed-optimization-based mix-flow scheduling mechanism for data center networks

  • ZHANG Tong1,2, REN Fengyuan3, SHU Ran4
Author information +
文章历史 +

摘要

数据中心网络作为云计算的关键基础设施，其性能对业务服务质量有至关重要的影响。在当前数据中心多业务并存的条件下，数据中心网络中同时存在截止期限流和非截止期限流。为同时满足2种流的传输需求，该文提出一种基于分布式优化的数据中心网络混流调度（distributed-optimization-based mix-flow scheduling，DOMS）机制。首先对截止期限流和非截止期限流分别定义优化目标和传输约束，将混流调度问题形式化为实时速率分配问题；然后利用问题的对偶分解特性，设计主机与交换机的协同调度结构，分布式求解该问题，设定每条流的传输速率并演化至全局最优解。仿真结果表明，DOMS能有效降低截止期限流的期限错失率和非截止期限流的完成时间。

Abstract

Data center networks are key cloud computing infrastructure whose performance critically impacts the quality of service. Currently, data centers have multiple services with both deadline and non-deadline flows. This paper presents a distributed-optimization-based mix-flow scheduling (DOMS) mechanism to meet the transmission requirements of both types of flows. First, the optimization goals and transmission constraints are defined for both kinds of flows and the mixed-flow scheduling problem is formalized as a real-time rate allocation problem. Then, a coordinated scheduling structure is designed for the hosts and switches that leverages the dual decomposition characteristics of the problem. This method uses a distributed solution method to solve the problem with the flow rates evolving to a global optimal solution. Simulations show that this method effectively reduces deadline miss rates for deadline flows as well as flow completion times for non-deadline flows.

关键词

数据中心网络 / 混流调度 / 分布式优化 / 截止期限错失率 / 流完成时间

Key words

data center network (DCN) / mix-flow scheduling / distributed optimization / deadline miss rate (DMR) / flow completion time (FCT)

引用本文

EndNote

Ris (Procite)

Bibtex

导出引用
张彤, 任丰原, 舒然. 基于分布式优化的数据中心网络混流调度机制[J]. 清华大学学报（自然科学版）. 2021, 61(6): 618-625 https://doi.org/10.16511/j.cnki.qhdxxb.2020.21.018
ZHANG Tong, REN Fengyuan, SHU Ran. Distributed-optimization-based mix-flow scheduling mechanism for data center networks[J]. Journal of Tsinghua University(Science and Technology). 2021, 61(6): 618-625 https://doi.org/10.16511/j.cnki.qhdxxb.2020.21.018

参考文献

[1] Cisco. Cisco global cloud index:Forecast and methodology, 2016-2021[R]. San Jose, USA:Cisco, 2018.
[2] NOORMOHAMMADPOUR M, RAGHAVENDRA C S. Datacenter traffic control:Understanding techniques and tradeoffs[J]. IEEE Communications Surveys and Tutorials, 2018, 20(2):1492-1525.
[3] CHEN L, CHEN K, BAI W, et al. Scheduling mix-flows in commodity datacenters with Karuna[C]//Proceedings of the ACM SIGCOMM 2016 Conference. Florianopolis, Brazil:ACM, 2016:174-187.
[4] WILSON C, BALLANI H, KARAGIANNIS T, et al. Better never than late:Meeting deadlines in datacenter networks[C]//Proceedings of the ACM SIGCOMM 2011 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications. Toronto, Canada:ACM, 2011:50-61.
[5] CHOWDHURY M, ZHONG Y, STOICA I. Efficient coflow scheduling with Varys[C]//ACM SIGCOMM 2014 Conference. Chicago, USA:ACM, 2014:443-454.
[6] HONG C Y, CAESAR M, GODFREY P B. Finishing flows quickly with preemptive scheduling[C]//ACM SIGCOMM 2012 Conference. Helsinki, Finland:ACM, 2012:127-138.
[7] ALIZADEH M, YANG S, SHARIF M, et al. pFabric:Minimal near-optimal datacenter transport[C]//ACM SIGCOMM 2013 Conference. Hong Kong, China:ACM, 2013:435-446.
[8] PERRY J, OUSTERHOUT A, BALAKRISHNAN H, et al. Fastpass:A centralized ‘zero-queue’ datacenter network[C]//ACM SIGCOMM 2014 Conference. Chicago, USA:ACM, 2014:307-318.
[9] MUNIR A, BAIG G, IRTEZA S M, et al. Friends, not foes:Synthesizing existing transport strategies for data center networks[C]//ACM SIGCOMM 2014 Conference. Chicago, USA:ACM, 2014:491-502.
[10] BAI W, CHEN L, CHEN K, et al. PIAS:Practical information-agnostic flow scheduling for commodity data centers[J]. IEEE/ACM Transactions on Networking, 2017, 25(4):1954-1967.
[11] GAO P X, NARAYAN A, KUMAR G, et al. pHost:Distributed near-optimal datacenter transport over commodity network fabric[C]//Proceedings of the 11th ACM Conference on Emerging Networking Experiments and Technologies. Heidelberg, Germany:ACM, 2015, 1:1-12.
[12] CHEN L, CHEN K, BAI W, et al. Scheduling mix-flows in commodity datacenters with Karuna[C]//Proceedings of the ACM SIGCOMM 2016 Conference. Florianopolis, Brazil:ACM, 2016:174-187.
[13] WANG T, XU H, LIU F M. Aemon:Information-agnostic mix-flow scheduling in data center networks[C]//Proceedings of the First Asia-Pacific Workshop on Networking. Hong Kong, China:ACM, 2017:106-112.
[14] 臧韦菲, 兰巨龙, 胡宇翔. 基于松弛时间与累计发送量的数据中心网络混合流调度机制[J]. 电子学报, 2019, 47(10):2061-2068.ZANG W F, LAN J L, HU Y X. Slack time and accumulation-based mix-flow scheduling in data center networks[J]. Chinese Journal of Electronics, 2019, 47(10):2061-2068. (in Chinese)
[15] ALIZADEH M, GREENBERG A G, DAVID A. M, et al. Data center TCP (DCTCP)[C]//Proceedings of the ACM SIGCOMM 2010 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications. New Delhi, India:ACM, 2010:63-74.
[16] GREENBERG A, HAMILTON J R, JAIN N, et al. VL2:A scalable and flexible data center network[C]//Proceedings of the ACM SIGCOMM 2009 Conference on Data Communication. Barcelona, Spain:ACM, 2009:51-62.

基金

国家自然科学基金面上项目（61872208）

PDF(1541 KB)

Accesses

Citation

Detail
段落导航
相关文章
扫码关注，本刊微信公众号
地址：清华大学东门外学研大厦B座605A室　
邮编：100084
电话：010-62788108/62771385/62792976/62792994　
E-mail：xuebaost@tsinghua.edu.cn
访问统计
    总访问量
    今日访问
    在线人数
　
版权所有 © 《清华大学学报（自然科学版）》编辑部

/

〈 〉