Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks

Shaojun Zou; Jiawei Huang; Jianxin Wang; Tian He

doi:10.1109/TNET.2020.3027749

Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks

Shaojun Zou, Jiawei Huang, Jianxin Wang, Tian He

Center for Transportation Studies

Research output: Contribution to journal › Article › peer-review

19 Scopus citations

Abstract

In data center networks, many network-intensive applications leverage large fan-in and many-to-one communication to achieve high performance. However, the special traffic patterns, such as micro-burst and high concurrency, easily cause TCP Incast problem and seriously degrade the application performance. To address the TCP Incast problem, we first reveal theoretically and empirically that alleviating packet burstiness is much more effective in reducing the Incast probability than controlling the congestion window. Inspired by the findings and insights from our experimental observations, we further propose a general supporting scheme Adaptive Pacing (AP), which dynamically adjusts burstiness according to the flow concurrency without any change on switch. Additionally, a sender-based approach is proposed to estimate the flow concurrency. Another feature of AP is its broad applicability. We integrate AP transparently into different TCP protocols (i.e., DCTCP, L2DCT and D2TCP). Through a series of large-scale NS2 simulations and testbed experiments, we show that AP significantly reduces the Incast probability across different TCP protocols and the network goodput can be increased consistently by on average 7times under severe congestion.

Original language	English (US)
Article number	9216628
Pages (from-to)	134-147
Number of pages	14
Journal	IEEE/ACM Transactions on Networking
Volume	29
Issue number	1
DOIs	https://doi.org/10.1109/TNET.2020.3027749
State	Published - Feb 2021

Bibliographical note

Funding Information:
Manuscript received October 12, 2017; revised September 30, 2018, December 7, 2019, March 27, 2020, and June 27, 2020; accepted September 21, 2020; approved by IEEE/ACM TRANSACTIONS ON NET-WORKING Editor Y. Guan. Date of publication October 7, 2020; date of current version February 17, 2021. This work was supported in part by the National Natural Science Foundation of China under Grant 61872387, Grant 61572530, and Grant 61872403; in part by the CERNET Innovation Project under Grant NGII20170107; and in part by the Fundamental Research Funds for Central Universities of Central South University under Grant 2017zzts142. A preliminary version of this article appears in IEEE ICDCS, Atlanta, GA, USA, July 2017. (Corresponding author: Jiawei Huang.) Shaojun Zou, Jiawei Huang, and Jianxin Wang are with the School of Computer Science and Engineering, Central South University, Changsha 410083, China (e-mail: jiaweihuang@csu.edu.cn).

Publisher Copyright:
© 1993-2012 IEEE.

Keywords

Data center
Incast
TCP
congestion control
pacing

Access

10.1109/TNET.2020.3027749

OpenUrl availability

Full text

Cite this

@article{7b976d76967a41ae99b484ad6d609b71,

title = "Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks",

abstract = "In data center networks, many network-intensive applications leverage large fan-in and many-to-one communication to achieve high performance. However, the special traffic patterns, such as micro-burst and high concurrency, easily cause TCP Incast problem and seriously degrade the application performance. To address the TCP Incast problem, we first reveal theoretically and empirically that alleviating packet burstiness is much more effective in reducing the Incast probability than controlling the congestion window. Inspired by the findings and insights from our experimental observations, we further propose a general supporting scheme Adaptive Pacing (AP), which dynamically adjusts burstiness according to the flow concurrency without any change on switch. Additionally, a sender-based approach is proposed to estimate the flow concurrency. Another feature of AP is its broad applicability. We integrate AP transparently into different TCP protocols (i.e., DCTCP, L2DCT and D2TCP). Through a series of large-scale NS2 simulations and testbed experiments, we show that AP significantly reduces the Incast probability across different TCP protocols and the network goodput can be increased consistently by on average 7times under severe congestion. ",

keywords = "Data center, Incast, TCP, congestion control, pacing",

author = "Shaojun Zou and Jiawei Huang and Jianxin Wang and Tian He",

note = "Funding Information: Manuscript received October 12, 2017; revised September 30, 2018, December 7, 2019, March 27, 2020, and June 27, 2020; accepted September 21, 2020; approved by IEEE/ACM TRANSACTIONS ON NET-WORKING Editor Y. Guan. Date of publication October 7, 2020; date of current version February 17, 2021. This work was supported in part by the National Natural Science Foundation of China under Grant 61872387, Grant 61572530, and Grant 61872403; in part by the CERNET Innovation Project under Grant NGII20170107; and in part by the Fundamental Research Funds for Central Universities of Central South University under Grant 2017zzts142. A preliminary version of this article appears in IEEE ICDCS, Atlanta, GA, USA, July 2017. (Corresponding author: Jiawei Huang.) Shaojun Zou, Jiawei Huang, and Jianxin Wang are with the School of Computer Science and Engineering, Central South University, Changsha 410083, China (e-mail: jiaweihuang@csu.edu.cn). Publisher Copyright: {\textcopyright} 1993-2012 IEEE.",

year = "2021",

month = feb,

doi = "10.1109/TNET.2020.3027749",

language = "English (US)",

volume = "29",

pages = "134--147",

journal = "IEEE/ACM Transactions on Networking",

issn = "1063-6692",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "1",

}

TY - JOUR

T1 - Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks

AU - Zou, Shaojun

AU - Huang, Jiawei

AU - Wang, Jianxin

AU - He, Tian

N1 - Funding Information: Manuscript received October 12, 2017; revised September 30, 2018, December 7, 2019, March 27, 2020, and June 27, 2020; accepted September 21, 2020; approved by IEEE/ACM TRANSACTIONS ON NET-WORKING Editor Y. Guan. Date of publication October 7, 2020; date of current version February 17, 2021. This work was supported in part by the National Natural Science Foundation of China under Grant 61872387, Grant 61572530, and Grant 61872403; in part by the CERNET Innovation Project under Grant NGII20170107; and in part by the Fundamental Research Funds for Central Universities of Central South University under Grant 2017zzts142. A preliminary version of this article appears in IEEE ICDCS, Atlanta, GA, USA, July 2017. (Corresponding author: Jiawei Huang.) Shaojun Zou, Jiawei Huang, and Jianxin Wang are with the School of Computer Science and Engineering, Central South University, Changsha 410083, China (e-mail: jiaweihuang@csu.edu.cn). Publisher Copyright: © 1993-2012 IEEE.

PY - 2021/2

Y1 - 2021/2

N2 - In data center networks, many network-intensive applications leverage large fan-in and many-to-one communication to achieve high performance. However, the special traffic patterns, such as micro-burst and high concurrency, easily cause TCP Incast problem and seriously degrade the application performance. To address the TCP Incast problem, we first reveal theoretically and empirically that alleviating packet burstiness is much more effective in reducing the Incast probability than controlling the congestion window. Inspired by the findings and insights from our experimental observations, we further propose a general supporting scheme Adaptive Pacing (AP), which dynamically adjusts burstiness according to the flow concurrency without any change on switch. Additionally, a sender-based approach is proposed to estimate the flow concurrency. Another feature of AP is its broad applicability. We integrate AP transparently into different TCP protocols (i.e., DCTCP, L2DCT and D2TCP). Through a series of large-scale NS2 simulations and testbed experiments, we show that AP significantly reduces the Incast probability across different TCP protocols and the network goodput can be increased consistently by on average 7times under severe congestion.

AB - In data center networks, many network-intensive applications leverage large fan-in and many-to-one communication to achieve high performance. However, the special traffic patterns, such as micro-burst and high concurrency, easily cause TCP Incast problem and seriously degrade the application performance. To address the TCP Incast problem, we first reveal theoretically and empirically that alleviating packet burstiness is much more effective in reducing the Incast probability than controlling the congestion window. Inspired by the findings and insights from our experimental observations, we further propose a general supporting scheme Adaptive Pacing (AP), which dynamically adjusts burstiness according to the flow concurrency without any change on switch. Additionally, a sender-based approach is proposed to estimate the flow concurrency. Another feature of AP is its broad applicability. We integrate AP transparently into different TCP protocols (i.e., DCTCP, L2DCT and D2TCP). Through a series of large-scale NS2 simulations and testbed experiments, we show that AP significantly reduces the Incast probability across different TCP protocols and the network goodput can be increased consistently by on average 7times under severe congestion.

KW - Data center

KW - Incast

KW - TCP

KW - congestion control

KW - pacing

UR - http://www.scopus.com/inward/record.url?scp=85092910839&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85092910839&partnerID=8YFLogxK

U2 - 10.1109/TNET.2020.3027749

DO - 10.1109/TNET.2020.3027749

M3 - Article

AN - SCOPUS:85092910839

SN - 1063-6692

VL - 29

SP - 134

EP - 147

JO - IEEE/ACM Transactions on Networking

JF - IEEE/ACM Transactions on Networking

IS - 1

M1 - 9216628

ER -

Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks

Abstract

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this