PABO: Mitigating Congestion via Packet Bounce in Data Center Networks

Xiang Shi, Lin Wang, Fa Zhang, Kai Zheng, Max Mühlhäuser, Zhiyong Liu

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

In today's data center, a diverse mix of throughput-sensitive long flows and delay-sensitive short flows are commonly presented. However, commodity switches used in a typical data center network are usually shallow-buffered for the sake of reducing queueing delay and deployment cost. The direct outcome is that the queue occupation by long flows could potentially block the transmission of delay-sensitive short flows, leading to degraded performance. Congestion can also be caused by the synchronization of multiple TCP connections for short flows, as typically seen in the partition/aggregate traffic pattern. The congestion is usually transient and any end-device intervention through the timeout-based pathway would result in suboptimal performance. While multiple end-to-end transport-layer solutions have been proposed, none of them has tackled the real challenge: reliable transmission in the network. In this paper, we fill this gap by presenting PABO — a novel link-layer design that can mitigate congestion by temporarily bouncing packets to upstream switches. PABO's design fulfills the following goals: (i) providing per-flow based flow control on the link layer, (ii) handling transient congestion without the intervention of end devices, and (iii) gradually back propagating the congestion signal to the source when the network is not capable to handle the congestion. We present the detailed design of PABO and complete a proof-of-concept implementation. We discuss the impact of system parameters on packet out-of-order delivery and conduct extensive experiments to prove the effectiveness of PABO. We examine the basic properties of PABO using a tree-based topology, and further evaluate the overall performance of PABO using a realistic Fattree topology for data center networks. Experiment results show that PABO can provide prominent advantage of mitigating transient congestions and can achieve significant gain on flow completion time.

LanguageEnglish
Pages1-14
Number of pages14
JournalComputer Communications
Volume140-141
Early online date16 Apr 2019
DOIs
Publication statusPublished - 1 May 2019

Fingerprint

Switches
Topology
Flow control
Synchronization
Experiments
Throughput
Costs

Cite this

Shi, Xiang ; Wang, Lin ; Zhang, Fa ; Zheng, Kai ; Mühlhäuser, Max ; Liu, Zhiyong. / PABO: Mitigating Congestion via Packet Bounce in Data Center Networks. In: Computer Communications. 2019 ; Vol. 140-141. pp. 1-14.
@article{d170490e989a4661a5494cd2344100a3,
title = "PABO: Mitigating Congestion via Packet Bounce in Data Center Networks",
abstract = "In today's data center, a diverse mix of throughput-sensitive long flows and delay-sensitive short flows are commonly presented. However, commodity switches used in a typical data center network are usually shallow-buffered for the sake of reducing queueing delay and deployment cost. The direct outcome is that the queue occupation by long flows could potentially block the transmission of delay-sensitive short flows, leading to degraded performance. Congestion can also be caused by the synchronization of multiple TCP connections for short flows, as typically seen in the partition/aggregate traffic pattern. The congestion is usually transient and any end-device intervention through the timeout-based pathway would result in suboptimal performance. While multiple end-to-end transport-layer solutions have been proposed, none of them has tackled the real challenge: reliable transmission in the network. In this paper, we fill this gap by presenting PABO — a novel link-layer design that can mitigate congestion by temporarily bouncing packets to upstream switches. PABO's design fulfills the following goals: (i) providing per-flow based flow control on the link layer, (ii) handling transient congestion without the intervention of end devices, and (iii) gradually back propagating the congestion signal to the source when the network is not capable to handle the congestion. We present the detailed design of PABO and complete a proof-of-concept implementation. We discuss the impact of system parameters on packet out-of-order delivery and conduct extensive experiments to prove the effectiveness of PABO. We examine the basic properties of PABO using a tree-based topology, and further evaluate the overall performance of PABO using a realistic Fattree topology for data center networks. Experiment results show that PABO can provide prominent advantage of mitigating transient congestions and can achieve significant gain on flow completion time.",
author = "Xiang Shi and Lin Wang and Fa Zhang and Kai Zheng and Max M{\"u}hlh{\"a}user and Zhiyong Liu",
year = "2019",
month = "5",
day = "1",
doi = "10.1016/j.comcom.2019.04.002",
language = "English",
volume = "140-141",
pages = "1--14",
journal = "Computer Communications",
issn = "0140-3664",
publisher = "Elsevier",

}

PABO: Mitigating Congestion via Packet Bounce in Data Center Networks. / Shi, Xiang; Wang, Lin; Zhang, Fa; Zheng, Kai; Mühlhäuser, Max; Liu, Zhiyong.

In: Computer Communications, Vol. 140-141, 01.05.2019, p. 1-14.

Research output: Contribution to JournalArticleAcademicpeer-review

TY - JOUR

T1 - PABO: Mitigating Congestion via Packet Bounce in Data Center Networks

AU - Shi, Xiang

AU - Wang, Lin

AU - Zhang, Fa

AU - Zheng, Kai

AU - Mühlhäuser, Max

AU - Liu, Zhiyong

PY - 2019/5/1

Y1 - 2019/5/1

N2 - In today's data center, a diverse mix of throughput-sensitive long flows and delay-sensitive short flows are commonly presented. However, commodity switches used in a typical data center network are usually shallow-buffered for the sake of reducing queueing delay and deployment cost. The direct outcome is that the queue occupation by long flows could potentially block the transmission of delay-sensitive short flows, leading to degraded performance. Congestion can also be caused by the synchronization of multiple TCP connections for short flows, as typically seen in the partition/aggregate traffic pattern. The congestion is usually transient and any end-device intervention through the timeout-based pathway would result in suboptimal performance. While multiple end-to-end transport-layer solutions have been proposed, none of them has tackled the real challenge: reliable transmission in the network. In this paper, we fill this gap by presenting PABO — a novel link-layer design that can mitigate congestion by temporarily bouncing packets to upstream switches. PABO's design fulfills the following goals: (i) providing per-flow based flow control on the link layer, (ii) handling transient congestion without the intervention of end devices, and (iii) gradually back propagating the congestion signal to the source when the network is not capable to handle the congestion. We present the detailed design of PABO and complete a proof-of-concept implementation. We discuss the impact of system parameters on packet out-of-order delivery and conduct extensive experiments to prove the effectiveness of PABO. We examine the basic properties of PABO using a tree-based topology, and further evaluate the overall performance of PABO using a realistic Fattree topology for data center networks. Experiment results show that PABO can provide prominent advantage of mitigating transient congestions and can achieve significant gain on flow completion time.

AB - In today's data center, a diverse mix of throughput-sensitive long flows and delay-sensitive short flows are commonly presented. However, commodity switches used in a typical data center network are usually shallow-buffered for the sake of reducing queueing delay and deployment cost. The direct outcome is that the queue occupation by long flows could potentially block the transmission of delay-sensitive short flows, leading to degraded performance. Congestion can also be caused by the synchronization of multiple TCP connections for short flows, as typically seen in the partition/aggregate traffic pattern. The congestion is usually transient and any end-device intervention through the timeout-based pathway would result in suboptimal performance. While multiple end-to-end transport-layer solutions have been proposed, none of them has tackled the real challenge: reliable transmission in the network. In this paper, we fill this gap by presenting PABO — a novel link-layer design that can mitigate congestion by temporarily bouncing packets to upstream switches. PABO's design fulfills the following goals: (i) providing per-flow based flow control on the link layer, (ii) handling transient congestion without the intervention of end devices, and (iii) gradually back propagating the congestion signal to the source when the network is not capable to handle the congestion. We present the detailed design of PABO and complete a proof-of-concept implementation. We discuss the impact of system parameters on packet out-of-order delivery and conduct extensive experiments to prove the effectiveness of PABO. We examine the basic properties of PABO using a tree-based topology, and further evaluate the overall performance of PABO using a realistic Fattree topology for data center networks. Experiment results show that PABO can provide prominent advantage of mitigating transient congestions and can achieve significant gain on flow completion time.

UR - http://www.scopus.com/inward/record.url?scp=85064439262&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064439262&partnerID=8YFLogxK

U2 - 10.1016/j.comcom.2019.04.002

DO - 10.1016/j.comcom.2019.04.002

M3 - Article

VL - 140-141

SP - 1

EP - 14

JO - Computer Communications

T2 - Computer Communications

JF - Computer Communications

SN - 0140-3664

ER -