HotCocoa

Hardware congestion control abstractions

Mina Tahmasbi Arashloo, Monia Ghobadi, Jennifer L. Rexford, David P. Walker

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

Congestion control in multi-tenant data centers is an active area of research because of its significant impact on customer experience, and, consequently, on revenue. Therefore, new algorithms and protocols are expected to emerge as the Cloud evolves. Deploying new congestion control algorithms in the end host's hypervisor allows frequent updates, but processing packets at high rates in the hypervisor and implementing the elements of a congestion control algorithm, such as traffic shapers and timestamps, in software have well-studied inaccuracies and CPU inefficiencies. In this paper, we argue for implementing the entire congestion control algorithm in programmable NICs. To do so, we identify the absence of hardware-aware programming abstractions as the most immediate challenge and solve it using a simple high-level domain specific language called HotCocoa. HotCocoa lies at a sweet spot between the ability to express a broad set of congestion control algorithms and efficient hardware implementation. It offers a set of hardware-aware COngestion COntrol Abstractions that enable operators to specify their algorithm without having to worry about low-level hardware primitives. To evaluate HotCocoa, we implement four congestion control algorithms (Reno, DCTCP, PCC, and TIMELY) and use simulations to show that HotCocoa's implementation of Reno perfectly tracks the behavior of a native implementation in C++.

Original languageEnglish (US)
Title of host publicationHotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks
PublisherAssociation for Computing Machinery, Inc
Pages108-114
Number of pages7
ISBN (Electronic)9781450355698
DOIs
StatePublished - Nov 30 2017
Event16th ACM Workshop on Hot Topics in Networks, HotNets 2017 - Palo Alto, United States
Duration: Nov 30 2017Dec 1 2017

Publication series

NameHotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks

Other

Other16th ACM Workshop on Hot Topics in Networks, HotNets 2017
CountryUnited States
CityPalo Alto
Period11/30/1712/1/17

Fingerprint

Hardware
Program processors
Mathematical operators
Network protocols
Processing

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications

Cite this

Arashloo, M. T., Ghobadi, M., Rexford, J. L., & Walker, D. P. (2017). HotCocoa: Hardware congestion control abstractions. In HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks (pp. 108-114). (HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks). Association for Computing Machinery, Inc. https://doi.org/10.1145/3152434.3152457
Arashloo, Mina Tahmasbi ; Ghobadi, Monia ; Rexford, Jennifer L. ; Walker, David P. / HotCocoa : Hardware congestion control abstractions. HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks. Association for Computing Machinery, Inc, 2017. pp. 108-114 (HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks).
@inproceedings{3aeca91d385944a0bf2fe084cd5b7f79,
title = "HotCocoa: Hardware congestion control abstractions",
abstract = "Congestion control in multi-tenant data centers is an active area of research because of its significant impact on customer experience, and, consequently, on revenue. Therefore, new algorithms and protocols are expected to emerge as the Cloud evolves. Deploying new congestion control algorithms in the end host's hypervisor allows frequent updates, but processing packets at high rates in the hypervisor and implementing the elements of a congestion control algorithm, such as traffic shapers and timestamps, in software have well-studied inaccuracies and CPU inefficiencies. In this paper, we argue for implementing the entire congestion control algorithm in programmable NICs. To do so, we identify the absence of hardware-aware programming abstractions as the most immediate challenge and solve it using a simple high-level domain specific language called HotCocoa. HotCocoa lies at a sweet spot between the ability to express a broad set of congestion control algorithms and efficient hardware implementation. It offers a set of hardware-aware COngestion COntrol Abstractions that enable operators to specify their algorithm without having to worry about low-level hardware primitives. To evaluate HotCocoa, we implement four congestion control algorithms (Reno, DCTCP, PCC, and TIMELY) and use simulations to show that HotCocoa's implementation of Reno perfectly tracks the behavior of a native implementation in C++.",
author = "Arashloo, {Mina Tahmasbi} and Monia Ghobadi and Rexford, {Jennifer L.} and Walker, {David P.}",
year = "2017",
month = "11",
day = "30",
doi = "10.1145/3152434.3152457",
language = "English (US)",
series = "HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks",
publisher = "Association for Computing Machinery, Inc",
pages = "108--114",
booktitle = "HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks",

}

Arashloo, MT, Ghobadi, M, Rexford, JL & Walker, DP 2017, HotCocoa: Hardware congestion control abstractions. in HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks. HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks, Association for Computing Machinery, Inc, pp. 108-114, 16th ACM Workshop on Hot Topics in Networks, HotNets 2017, Palo Alto, United States, 11/30/17. https://doi.org/10.1145/3152434.3152457

HotCocoa : Hardware congestion control abstractions. / Arashloo, Mina Tahmasbi; Ghobadi, Monia; Rexford, Jennifer L.; Walker, David P.

HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks. Association for Computing Machinery, Inc, 2017. p. 108-114 (HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - HotCocoa

T2 - Hardware congestion control abstractions

AU - Arashloo, Mina Tahmasbi

AU - Ghobadi, Monia

AU - Rexford, Jennifer L.

AU - Walker, David P.

PY - 2017/11/30

Y1 - 2017/11/30

N2 - Congestion control in multi-tenant data centers is an active area of research because of its significant impact on customer experience, and, consequently, on revenue. Therefore, new algorithms and protocols are expected to emerge as the Cloud evolves. Deploying new congestion control algorithms in the end host's hypervisor allows frequent updates, but processing packets at high rates in the hypervisor and implementing the elements of a congestion control algorithm, such as traffic shapers and timestamps, in software have well-studied inaccuracies and CPU inefficiencies. In this paper, we argue for implementing the entire congestion control algorithm in programmable NICs. To do so, we identify the absence of hardware-aware programming abstractions as the most immediate challenge and solve it using a simple high-level domain specific language called HotCocoa. HotCocoa lies at a sweet spot between the ability to express a broad set of congestion control algorithms and efficient hardware implementation. It offers a set of hardware-aware COngestion COntrol Abstractions that enable operators to specify their algorithm without having to worry about low-level hardware primitives. To evaluate HotCocoa, we implement four congestion control algorithms (Reno, DCTCP, PCC, and TIMELY) and use simulations to show that HotCocoa's implementation of Reno perfectly tracks the behavior of a native implementation in C++.

AB - Congestion control in multi-tenant data centers is an active area of research because of its significant impact on customer experience, and, consequently, on revenue. Therefore, new algorithms and protocols are expected to emerge as the Cloud evolves. Deploying new congestion control algorithms in the end host's hypervisor allows frequent updates, but processing packets at high rates in the hypervisor and implementing the elements of a congestion control algorithm, such as traffic shapers and timestamps, in software have well-studied inaccuracies and CPU inefficiencies. In this paper, we argue for implementing the entire congestion control algorithm in programmable NICs. To do so, we identify the absence of hardware-aware programming abstractions as the most immediate challenge and solve it using a simple high-level domain specific language called HotCocoa. HotCocoa lies at a sweet spot between the ability to express a broad set of congestion control algorithms and efficient hardware implementation. It offers a set of hardware-aware COngestion COntrol Abstractions that enable operators to specify their algorithm without having to worry about low-level hardware primitives. To evaluate HotCocoa, we implement four congestion control algorithms (Reno, DCTCP, PCC, and TIMELY) and use simulations to show that HotCocoa's implementation of Reno perfectly tracks the behavior of a native implementation in C++.

UR - http://www.scopus.com/inward/record.url?scp=85041200366&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85041200366&partnerID=8YFLogxK

U2 - 10.1145/3152434.3152457

DO - 10.1145/3152434.3152457

M3 - Conference contribution

T3 - HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks

SP - 108

EP - 114

BT - HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks

PB - Association for Computing Machinery, Inc

ER -

Arashloo MT, Ghobadi M, Rexford JL, Walker DP. HotCocoa: Hardware congestion control abstractions. In HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks. Association for Computing Machinery, Inc. 2017. p. 108-114. (HotNets 2017 - Proceedings of the 16th ACM Workshop on Hot Topics in Networks). https://doi.org/10.1145/3152434.3152457