On the performance of scalable fast optical switch based recursive network architecture for HPC applications

Fulong Yan; Elham Khani; Xiaotao Guo; Xuwei Xue; Bitao Pan; Nicola Calabretta

doi:10.1049/cp.2019.0932

On the performance of scalable fast optical switch based recursive network architecture for HPC applications

Fulong Yan, Elham Khani, Xiaotao Guo, Xuwei Xue, Bitao Pan, Nicola Calabretta

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/Congresprocedure › Conferentiebijdrage › Academic › peer review

1 Citaat (Scopus)

1 Downloads (Pure)

Samenvatting

We investigate a novel scalable low-latency HPC network for blade-to-blade interconnection exploiting multiple-parallel recursive architecture with fast optical switches under realistic HPC application traffics. Compared with Leaf-Spine HPC interconnecting 4096-blade, results show that HFOS4 reduces up to 75.9% latency, 64.2% cost and 30.9% power consumption.

Originele taal-2	Engels
Titel	45th European Conference on Optical Communication (ECOC 2019)
Uitgeverij	Institution of Engineering and Technology (IET)
Aantal pagina's	4
ISBN van geprinte versie	978-1-83953-185-9
DOI's	https://doi.org/10.1049/cp.2019.0932
Status	Gepubliceerd - 30 jun. 2020
Evenement	45th European Conference on Optical Communication (ECOC 2019) - Dublin, Ierland Duur: 22 sep. 2019 → 26 sep. 2019 Congresnummer: 45 https://www.ecoc2019.org/

Publicatie series

Naam	IET Conference Publications
Nummer	CP765

Congres

Congres	45th European Conference on Optical Communication (ECOC 2019)
Verkorte titel	ECOC 2019
Land/Regio	Ierland
Stad	Dublin
Periode	22/09/19 → 26/09/19
Internet adres	https://www.ecoc2019.org/

Toegang tot document

10.1049/cp.2019.0932

Andere bestanden en links

Link to publication in Scopus

Citeer dit

Yan, F., Khani, E., Guo, X., Xue, X., Pan, B., & Calabretta, N. (2020). On the performance of scalable fast optical switch based recursive network architecture for HPC applications. In 45th European Conference on Optical Communication (ECOC 2019) [9125660] (IET Conference Publications; Nr. CP765). Institution of Engineering and Technology (IET). https://doi.org/10.1049/cp.2019.0932

@inproceedings{1bb73abb70964a468ab3ba3c13335d40,

title = "On the performance of scalable fast optical switch based recursive network architecture for HPC applications",

abstract = "We investigate a novel scalable low-latency HPC network for blade-to-blade interconnection exploiting multiple-parallel recursive architecture with fast optical switches under realistic HPC application traffics. Compared with Leaf-Spine HPC interconnecting 4096-blade, results show that HFOS4 reduces up to 75.9% latency, 64.2% cost and 30.9% power consumption.",

keywords = "High performance computing, Network architecture, Optical switches, System performance",

author = "Fulong Yan and Elham Khani and Xiaotao Guo and Xuwei Xue and Bitao Pan and Nicola Calabretta",

year = "2020",

month = jun,

day = "30",

doi = "10.1049/cp.2019.0932",

language = "English",

isbn = "978-1-83953-185-9",

series = "IET Conference Publications",

publisher = "Institution of Engineering and Technology (IET)",

number = "CP765",

booktitle = "45th European Conference on Optical Communication (ECOC 2019)",

address = "United Kingdom",

note = "45th European Conference on Optical Communication (ECOC 2019), ECOC 2019 ; Conference date: 22-09-2019 Through 26-09-2019",

url = "https://www.ecoc2019.org/",

}

Yan, F , Khani, E, Guo, X, Xue, X , Pan, B & Calabretta, N 2020, On the performance of scalable fast optical switch based recursive network architecture for HPC applications. in 45th European Conference on Optical Communication (ECOC 2019)., 9125660, IET Conference Publications, nr. CP765, Institution of Engineering and Technology (IET), 45th European Conference on Optical Communication (ECOC 2019), Dublin, Ierland, 22/09/19. https://doi.org/10.1049/cp.2019.0932

On the performance of scalable fast optical switch based recursive network architecture for HPC applications. / Yan, Fulong ; Khani, Elham; Guo, Xiaotao et al.

45th European Conference on Optical Communication (ECOC 2019). Institution of Engineering and Technology (IET), 2020. 9125660 (IET Conference Publications; Nr. CP765).

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/Congresprocedure › Conferentiebijdrage › Academic › peer review

TY - GEN

T1 - On the performance of scalable fast optical switch based recursive network architecture for HPC applications

AU - Yan, Fulong

AU - Khani, Elham

AU - Guo, Xiaotao

AU - Xue, Xuwei

AU - Pan, Bitao

AU - Calabretta, Nicola

N1 - Conference code: 45

PY - 2020/6/30

Y1 - 2020/6/30

N2 - We investigate a novel scalable low-latency HPC network for blade-to-blade interconnection exploiting multiple-parallel recursive architecture with fast optical switches under realistic HPC application traffics. Compared with Leaf-Spine HPC interconnecting 4096-blade, results show that HFOS4 reduces up to 75.9% latency, 64.2% cost and 30.9% power consumption.

AB - We investigate a novel scalable low-latency HPC network for blade-to-blade interconnection exploiting multiple-parallel recursive architecture with fast optical switches under realistic HPC application traffics. Compared with Leaf-Spine HPC interconnecting 4096-blade, results show that HFOS4 reduces up to 75.9% latency, 64.2% cost and 30.9% power consumption.

KW - High performance computing

KW - Network architecture

KW - Optical switches

KW - System performance

UR - http://www.scopus.com/inward/record.url?scp=85086900052&partnerID=8YFLogxK

U2 - 10.1049/cp.2019.0932

DO - 10.1049/cp.2019.0932

M3 - Conference contribution

AN - SCOPUS:85086900052

SN - 978-1-83953-185-9

T3 - IET Conference Publications

BT - 45th European Conference on Optical Communication (ECOC 2019)

PB - Institution of Engineering and Technology (IET)

T2 - 45th European Conference on Optical Communication (ECOC 2019)

Y2 - 22 September 2019 through 26 September 2019

ER -

On the performance of scalable fast optical switch based recursive network architecture for HPC applications

Samenvatting

Publicatie series

Congres

Toegang tot document

Andere bestanden en links

Vingerafdruk

Citeer dit