Absorbing Microbursts Without Headroom for Data Center Networks

Gyuyeong Kim; Wonjun Lee

doi:10.1109/LCOMM.2019.2907087

Absorbing Microbursts Without Headroom for Data Center Networks

Gyuyeong Kim, Wonjun Lee

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

7 Citations (Scopus)

Abstract

In data center networks, handling microbursts is crucial to user experience because the microburst causes excessive packet losses in shallow buffered switches. Explicit congestion notification (ECN) enables the switch to absorb microbursts by leaving buffer headroom, but has a fundamental trade-off between latency and throughput. In this letter, we propose LossPass, a switch port management scheme that achieves low latency and high throughput simultaneously. LossPass breaks the trade-off of ECN by absorbing microbursts without headroom. Intensive benchmark results demonstrate that LossPass reduces the average and tail latency of small flows while maintaining line-rate throughput.

Original language	English
Article number	8672925
Pages (from-to)	806-809
Number of pages	4
Journal	IEEE Communications Letters
Volume	23
Issue number	5
DOIs	https://doi.org/10.1109/LCOMM.2019.2907087
Publication status	Published - 2019 May

Bibliographical note

Funding Information:
Manuscript received December 19, 2018; revised February 11, 2019; accepted March 19, 2019. Date of publication March 22, 2019; date of current version May 8, 2019. This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676. The associate editor coordinating the review of this letter and approving it for publication was A. Al-Fuqaha. (Corresponding author: Wonjun Lee.) The authors are with the Network and Security Research Laboratory, School of Cybersecurity, Korea University, Seoul 02841, South Korea (e-mail: [email protected]; [email protected]). Digital Object Identifier 10.1109/LCOMM.2019.2907087 Fig. 1. The trade-off between latency and throughput in ECN. Through ns-2, we simulate a compute rack consisting of 32 servers and a ToR switch. Overall, we generate 10K queries where each of 32 senders responds with 32KB of data. C = 10Gbps,RT T = 100µs, BDP ≈ 84pkts.

Funding Information:
This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676.

Publisher Copyright:
© 2019 IEEE.

Keywords

Data center networks
ECN
microbursts

ASJC Scopus subject areas

Modelling and Simulation
Computer Science Applications
Electrical and Electronic Engineering

Access to Document

10.1109/LCOMM.2019.2907087

Cite this

@article{8d713c96ccc24d41b124211488e3da77,

title = "Absorbing Microbursts Without Headroom for Data Center Networks",

abstract = "In data center networks, handling microbursts is crucial to user experience because the microburst causes excessive packet losses in shallow buffered switches. Explicit congestion notification (ECN) enables the switch to absorb microbursts by leaving buffer headroom, but has a fundamental trade-off between latency and throughput. In this letter, we propose LossPass, a switch port management scheme that achieves low latency and high throughput simultaneously. LossPass breaks the trade-off of ECN by absorbing microbursts without headroom. Intensive benchmark results demonstrate that LossPass reduces the average and tail latency of small flows while maintaining line-rate throughput.",

keywords = "Data center networks, ECN, microbursts",

author = "Gyuyeong Kim and Wonjun Lee",

note = "Funding Information: Manuscript received December 19, 2018; revised February 11, 2019; accepted March 19, 2019. Date of publication March 22, 2019; date of current version May 8, 2019. This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676. The associate editor coordinating the review of this letter and approving it for publication was A. Al-Fuqaha. (Corresponding author: Wonjun Lee.) The authors are with the Network and Security Research Laboratory, School of Cybersecurity, Korea University, Seoul 02841, South Korea (e-mail: [email protected]; [email protected]). Digital Object Identifier 10.1109/LCOMM.2019.2907087 Fig. 1. The trade-off between latency and throughput in ECN. Through ns-2, we simulate a compute rack consisting of 32 servers and a ToR switch. Overall, we generate 10K queries where each of 32 senders responds with 32KB of data. C = 10Gbps,RT T = 100µs, BDP ≈ 84pkts. Funding Information: This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676. Publisher Copyright: {\textcopyright} 2019 IEEE.",

year = "2019",

month = may,

doi = "10.1109/LCOMM.2019.2907087",

language = "English",

volume = "23",

pages = "806--809",

journal = "IEEE Communications Letters",

issn = "1089-7798",

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

number = "5",

}

TY - JOUR

T1 - Absorbing Microbursts Without Headroom for Data Center Networks

AU - Kim, Gyuyeong

AU - Lee, Wonjun

N1 - Funding Information: Manuscript received December 19, 2018; revised February 11, 2019; accepted March 19, 2019. Date of publication March 22, 2019; date of current version May 8, 2019. This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676. The associate editor coordinating the review of this letter and approving it for publication was A. Al-Fuqaha. (Corresponding author: Wonjun Lee.) The authors are with the Network and Security Research Laboratory, School of Cybersecurity, Korea University, Seoul 02841, South Korea (e-mail: [email protected]; [email protected]). Digital Object Identifier 10.1109/LCOMM.2019.2907087 Fig. 1. The trade-off between latency and throughput in ECN. Through ns-2, we simulate a compute rack consisting of 32 servers and a ToR switch. Overall, we generate 10K queries where each of 32 senders responds with 32KB of data. C = 10Gbps,RT T = 100µs, BDP ≈ 84pkts. Funding Information: This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and Information and Communication Technology (ICT) under Grant 2017R1A2B2004811, and in part by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT under Grant 2017M3C4A7083676. Publisher Copyright: © 2019 IEEE.

PY - 2019/5

Y1 - 2019/5

N2 - In data center networks, handling microbursts is crucial to user experience because the microburst causes excessive packet losses in shallow buffered switches. Explicit congestion notification (ECN) enables the switch to absorb microbursts by leaving buffer headroom, but has a fundamental trade-off between latency and throughput. In this letter, we propose LossPass, a switch port management scheme that achieves low latency and high throughput simultaneously. LossPass breaks the trade-off of ECN by absorbing microbursts without headroom. Intensive benchmark results demonstrate that LossPass reduces the average and tail latency of small flows while maintaining line-rate throughput.

AB - In data center networks, handling microbursts is crucial to user experience because the microburst causes excessive packet losses in shallow buffered switches. Explicit congestion notification (ECN) enables the switch to absorb microbursts by leaving buffer headroom, but has a fundamental trade-off between latency and throughput. In this letter, we propose LossPass, a switch port management scheme that achieves low latency and high throughput simultaneously. LossPass breaks the trade-off of ECN by absorbing microbursts without headroom. Intensive benchmark results demonstrate that LossPass reduces the average and tail latency of small flows while maintaining line-rate throughput.

KW - Data center networks

KW - ECN

KW - microbursts

UR - http://www.scopus.com/inward/record.url?scp=85065595083&partnerID=8YFLogxK

U2 - 10.1109/LCOMM.2019.2907087

DO - 10.1109/LCOMM.2019.2907087

M3 - Article

AN - SCOPUS:85065595083

SN - 1089-7798

VL - 23

SP - 806

EP - 809

JO - IEEE Communications Letters

JF - IEEE Communications Letters

IS - 5

M1 - 8672925

ER -

Absorbing Microbursts Without Headroom for Data Center Networks

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this