Dual-centric data center network architectures

Dawei Li, Jie Wu, Zhiyong Liu, Fa Zhang

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

6 Citations (Scopus)

Abstract

Existing Data Center Network (DCN) architectures are classified into two categories: switch-centric and server-centric architectures. In switch-centric DCNs, routing intelligence is placed on switches, each server usually uses only one port of the Network Interface Card (NIC) to connect to the network. In server-centric DCNs, switches are only used as cross-bars, and routing intelligence is placed on servers, where multiple NIC ports may be used. In this paper, we formally introduce a new category of DCN architectures: the dual-centric DCN architectures, where routing intelligence can be placed on both switches and servers. We propose two typical dual-centric DCN architectures: FSquare and Rectangle, both of which are based on the folded Clos topology. FSquare is a high performance DCN architecture, in which the diameter is small and the bisection bandwidth is large, however, the DCN power consumption per server in FSquare is high. Rectangle significantly reduces the DCN power consumption per server, compared to FSquare, at the sacrifice of some performances, thus, Rectangle has a larger diameter and a smaller bisection bandwidth. By investigating FSquare and Rectangle, and by comparing them with existing architectures, we demonstrate that, these two novel dual-centric architectures enjoy the advantages of both switch-centric designs and server-centric designs, have various nice properties for practical data centers, and provide flexible choices in designing DCN architectures.

Original languageEnglish
Title of host publicationProceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages679-688
Number of pages10
ISBN (Electronic)9781467375870
DOIs
StatePublished - 8 Dec 2015
Event44th International Conference on Parallel Processing, ICPP 2015 - Beijing, China
Duration: 1 Sep 20154 Sep 2015

Publication series

NameProceedings of the International Conference on Parallel Processing
Volume2015-December
ISSN (Print)0190-3918

Other

Other44th International Conference on Parallel Processing, ICPP 2015
CountryChina
CityBeijing
Period1/09/154/09/15

Fingerprint

Data Center
Network Architecture
Network architecture
Servers
Server
Switch
Switches
Rectangle
Routing
Bisection
Interfaces (computer)
Power Consumption
Electric power utilization
Bandwidth
Topology
High Performance

Keywords

  • Bisection bandwidth
  • Data center network (DCN)
  • Dual-centric design
  • End-to-end-delay
  • Power efficiency

Cite this

Li, D., Wu, J., Liu, Z., & Zhang, F. (2015). Dual-centric data center network architectures. In Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015 (pp. 679-688). [7349623] (Proceedings of the International Conference on Parallel Processing; Vol. 2015-December). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICPP.2015.77
Li, Dawei ; Wu, Jie ; Liu, Zhiyong ; Zhang, Fa. / Dual-centric data center network architectures. Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 679-688 (Proceedings of the International Conference on Parallel Processing).
@inproceedings{b896931f88c64281ba62106609f2487c,
title = "Dual-centric data center network architectures",
abstract = "Existing Data Center Network (DCN) architectures are classified into two categories: switch-centric and server-centric architectures. In switch-centric DCNs, routing intelligence is placed on switches, each server usually uses only one port of the Network Interface Card (NIC) to connect to the network. In server-centric DCNs, switches are only used as cross-bars, and routing intelligence is placed on servers, where multiple NIC ports may be used. In this paper, we formally introduce a new category of DCN architectures: the dual-centric DCN architectures, where routing intelligence can be placed on both switches and servers. We propose two typical dual-centric DCN architectures: FSquare and Rectangle, both of which are based on the folded Clos topology. FSquare is a high performance DCN architecture, in which the diameter is small and the bisection bandwidth is large, however, the DCN power consumption per server in FSquare is high. Rectangle significantly reduces the DCN power consumption per server, compared to FSquare, at the sacrifice of some performances, thus, Rectangle has a larger diameter and a smaller bisection bandwidth. By investigating FSquare and Rectangle, and by comparing them with existing architectures, we demonstrate that, these two novel dual-centric architectures enjoy the advantages of both switch-centric designs and server-centric designs, have various nice properties for practical data centers, and provide flexible choices in designing DCN architectures.",
keywords = "Bisection bandwidth, Data center network (DCN), Dual-centric design, End-to-end-delay, Power efficiency",
author = "Dawei Li and Jie Wu and Zhiyong Liu and Fa Zhang",
year = "2015",
month = "12",
day = "8",
doi = "10.1109/ICPP.2015.77",
language = "English",
series = "Proceedings of the International Conference on Parallel Processing",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "679--688",
booktitle = "Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015",

}

Li, D, Wu, J, Liu, Z & Zhang, F 2015, Dual-centric data center network architectures. in Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015., 7349623, Proceedings of the International Conference on Parallel Processing, vol. 2015-December, Institute of Electrical and Electronics Engineers Inc., pp. 679-688, 44th International Conference on Parallel Processing, ICPP 2015, Beijing, China, 1/09/15. https://doi.org/10.1109/ICPP.2015.77

Dual-centric data center network architectures. / Li, Dawei; Wu, Jie; Liu, Zhiyong; Zhang, Fa.

Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015. Institute of Electrical and Electronics Engineers Inc., 2015. p. 679-688 7349623 (Proceedings of the International Conference on Parallel Processing; Vol. 2015-December).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - Dual-centric data center network architectures

AU - Li, Dawei

AU - Wu, Jie

AU - Liu, Zhiyong

AU - Zhang, Fa

PY - 2015/12/8

Y1 - 2015/12/8

N2 - Existing Data Center Network (DCN) architectures are classified into two categories: switch-centric and server-centric architectures. In switch-centric DCNs, routing intelligence is placed on switches, each server usually uses only one port of the Network Interface Card (NIC) to connect to the network. In server-centric DCNs, switches are only used as cross-bars, and routing intelligence is placed on servers, where multiple NIC ports may be used. In this paper, we formally introduce a new category of DCN architectures: the dual-centric DCN architectures, where routing intelligence can be placed on both switches and servers. We propose two typical dual-centric DCN architectures: FSquare and Rectangle, both of which are based on the folded Clos topology. FSquare is a high performance DCN architecture, in which the diameter is small and the bisection bandwidth is large, however, the DCN power consumption per server in FSquare is high. Rectangle significantly reduces the DCN power consumption per server, compared to FSquare, at the sacrifice of some performances, thus, Rectangle has a larger diameter and a smaller bisection bandwidth. By investigating FSquare and Rectangle, and by comparing them with existing architectures, we demonstrate that, these two novel dual-centric architectures enjoy the advantages of both switch-centric designs and server-centric designs, have various nice properties for practical data centers, and provide flexible choices in designing DCN architectures.

AB - Existing Data Center Network (DCN) architectures are classified into two categories: switch-centric and server-centric architectures. In switch-centric DCNs, routing intelligence is placed on switches, each server usually uses only one port of the Network Interface Card (NIC) to connect to the network. In server-centric DCNs, switches are only used as cross-bars, and routing intelligence is placed on servers, where multiple NIC ports may be used. In this paper, we formally introduce a new category of DCN architectures: the dual-centric DCN architectures, where routing intelligence can be placed on both switches and servers. We propose two typical dual-centric DCN architectures: FSquare and Rectangle, both of which are based on the folded Clos topology. FSquare is a high performance DCN architecture, in which the diameter is small and the bisection bandwidth is large, however, the DCN power consumption per server in FSquare is high. Rectangle significantly reduces the DCN power consumption per server, compared to FSquare, at the sacrifice of some performances, thus, Rectangle has a larger diameter and a smaller bisection bandwidth. By investigating FSquare and Rectangle, and by comparing them with existing architectures, we demonstrate that, these two novel dual-centric architectures enjoy the advantages of both switch-centric designs and server-centric designs, have various nice properties for practical data centers, and provide flexible choices in designing DCN architectures.

KW - Bisection bandwidth

KW - Data center network (DCN)

KW - Dual-centric design

KW - End-to-end-delay

KW - Power efficiency

UR - http://www.scopus.com/inward/record.url?scp=84976499530&partnerID=8YFLogxK

U2 - 10.1109/ICPP.2015.77

DO - 10.1109/ICPP.2015.77

M3 - Conference contribution

T3 - Proceedings of the International Conference on Parallel Processing

SP - 679

EP - 688

BT - Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Li D, Wu J, Liu Z, Zhang F. Dual-centric data center network architectures. In Proceedings - 2015 44th International Annual Conference on Parallel Processing, ICPP 2015. Institute of Electrical and Electronics Engineers Inc. 2015. p. 679-688. 7349623. (Proceedings of the International Conference on Parallel Processing). https://doi.org/10.1109/ICPP.2015.77