TCP flow classification and bandwidth aggregation in optically interconnected data center networks

Houman Rastegarfar, Madeleine Glick, Nicolaas Viljoen, Mingwei Yang, John W Wissinger, Lloyd J LaComb, Nasser N Peyghambarian

Research output: Contribution to journalArticle

20 Citations (Scopus)

Abstract

Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidthhungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive- increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5% improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7%.

Original languageEnglish (US)
Article number7588243
Pages (from-to)777-786
Number of pages10
JournalJournal of Optical Communications and Networking
Volume8
Issue number10
DOIs
StatePublished - 2016

Fingerprint

Agglomeration
Bandwidth
Throughput
Transmission control protocol
Networks (circuits)
Channel capacity
Network performance
Telecommunication traffic
Consolidation
Tuning
Simulators
Modulation
Planning
Wavelength
Communication
Processing

Keywords

  • Bandwidth aggregation
  • Congestion control
  • Data center
  • Elephant flow
  • Flow classification
  • Machine learning
  • Mouse flow
  • TCP protocol

ASJC Scopus subject areas

  • Computer Networks and Communications

Cite this

TCP flow classification and bandwidth aggregation in optically interconnected data center networks. / Rastegarfar, Houman; Glick, Madeleine; Viljoen, Nicolaas; Yang, Mingwei; Wissinger, John W; LaComb, Lloyd J; Peyghambarian, Nasser N.

In: Journal of Optical Communications and Networking, Vol. 8, No. 10, 7588243, 2016, p. 777-786.

Research output: Contribution to journalArticle

@article{498e55b9154c4c6ca1972901793ef661,
title = "TCP flow classification and bandwidth aggregation in optically interconnected data center networks",
abstract = "Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidthhungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive- increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5{\%} improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7{\%}.",
keywords = "Bandwidth aggregation, Congestion control, Data center, Elephant flow, Flow classification, Machine learning, Mouse flow, TCP protocol",
author = "Houman Rastegarfar and Madeleine Glick and Nicolaas Viljoen and Mingwei Yang and Wissinger, {John W} and LaComb, {Lloyd J} and Peyghambarian, {Nasser N}",
year = "2016",
doi = "10.1364/JOCN.8.000777",
language = "English (US)",
volume = "8",
pages = "777--786",
journal = "Journal of Optical Communications and Networking",
issn = "1943-0620",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "10",

}

TY - JOUR

T1 - TCP flow classification and bandwidth aggregation in optically interconnected data center networks

AU - Rastegarfar, Houman

AU - Glick, Madeleine

AU - Viljoen, Nicolaas

AU - Yang, Mingwei

AU - Wissinger, John W

AU - LaComb, Lloyd J

AU - Peyghambarian, Nasser N

PY - 2016

Y1 - 2016

N2 - Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidthhungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive- increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5% improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7%.

AB - Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidthhungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive- increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5% improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7%.

KW - Bandwidth aggregation

KW - Congestion control

KW - Data center

KW - Elephant flow

KW - Flow classification

KW - Machine learning

KW - Mouse flow

KW - TCP protocol

UR - http://www.scopus.com/inward/record.url?scp=84990943740&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990943740&partnerID=8YFLogxK

U2 - 10.1364/JOCN.8.000777

DO - 10.1364/JOCN.8.000777

M3 - Article

AN - SCOPUS:84990943740

VL - 8

SP - 777

EP - 786

JO - Journal of Optical Communications and Networking

JF - Journal of Optical Communications and Networking

SN - 1943-0620

IS - 10

M1 - 7588243

ER -