The Case for Network Coding for Collective Communication on HPC Interconnection Networks

Ahmed SHALABY; Ikki FUJIWARA; Michihiro KOIBUCHI

doi:10.1587/transinf.2014EDP7255

Open Access
The Case for Network Coding for Collective Communication on HPC Interconnection Networks

Ahmed SHALABY, Ikki FUJIWARA, Michihiro KOIBUCHI

Full Text Views

92

Cite this

Free PDF (1.7MB)

Summary :

Recently network bandwidth becomes a performance concern particularly for collective communication since bisection bandwidths of supercomputers become far less than their full bisection bandwidths. In this context we propose the use of a network coding technique to reduce the number of unicasts and the size of data transferred in latency-sensitive collective communications in supercomputers. Our proposed network coding scheme has a hierarchical multicasting structure with intra-group and inter-group unicasts. Quantitative analysis show that the aggregate path hop counts by our hierarchical network coding decrease as much as 94% when compared to conventional unicast-based multicasts. We validate these results by cycle-accurate network simulations. In 1,024-switch networks, the network reduces the execution time of collective communications as much as 70%. We also show that our hierarchical network coding is beneficial for any packet size.

Publication: IEICE TRANSACTIONS on Information Vol.E98-D No.3 pp.661-670

Publication Date: 2015/03/01

Publicized: 2014/12/11

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2014EDP7255

Type of Manuscript: PAPER

Category: Information Network

Authors

Ahmed SHALABY
  E-JUST
Ikki FUJIWARA
  National Institute of Informatics
Michihiro KOIBUCHI
  National Institute of Informatics

Keyword

interconnection networks, collective communication, network coding, high-performance computing

Cite this

Copy

Ahmed SHALABY, Ikki FUJIWARA, Michihiro KOIBUCHI, "The Case for Network Coding for Collective Communication on HPC Interconnection Networks" in IEICE TRANSACTIONS on Information, vol. E98-D, no. 3, pp. 661-670, March 2015, doi: 10.1587/transinf.2014EDP7255.
Abstract: Recently network bandwidth becomes a performance concern particularly for collective communication since bisection bandwidths of supercomputers become far less than their full bisection bandwidths. In this context we propose the use of a network coding technique to reduce the number of unicasts and the size of data transferred in latency-sensitive collective communications in supercomputers. Our proposed network coding scheme has a hierarchical multicasting structure with intra-group and inter-group unicasts. Quantitative analysis show that the aggregate path hop counts by our hierarchical network coding decrease as much as 94% when compared to conventional unicast-based multicasts. We validate these results by cycle-accurate network simulations. In 1,024-switch networks, the network reduces the execution time of collective communications as much as 70%. We also show that our hierarchical network coding is beneficial for any packet size.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2014EDP7255/_p

Copy

@ARTICLE{e98-d_3_661,
author={Ahmed SHALABY, Ikki FUJIWARA, Michihiro KOIBUCHI, },
journal={IEICE TRANSACTIONS on Information},
title={The Case for Network Coding for Collective Communication on HPC Interconnection Networks},
year={2015},
volume={E98-D},
number={3},
pages={661-670},
abstract={Recently network bandwidth becomes a performance concern particularly for collective communication since bisection bandwidths of supercomputers become far less than their full bisection bandwidths. In this context we propose the use of a network coding technique to reduce the number of unicasts and the size of data transferred in latency-sensitive collective communications in supercomputers. Our proposed network coding scheme has a hierarchical multicasting structure with intra-group and inter-group unicasts. Quantitative analysis show that the aggregate path hop counts by our hierarchical network coding decrease as much as 94% when compared to conventional unicast-based multicasts. We validate these results by cycle-accurate network simulations. In 1,024-switch networks, the network reduces the execution time of collective communications as much as 70%. We also show that our hierarchical network coding is beneficial for any packet size.},
keywords={},
doi={10.1587/transinf.2014EDP7255},
ISSN={1745-1361},
month={March},}

Copy

TY - JOUR
TI - The Case for Network Coding for Collective Communication on HPC Interconnection Networks
T2 - IEICE TRANSACTIONS on Information
SP - 661
EP - 670
AU - Ahmed SHALABY
AU - Ikki FUJIWARA
AU - Michihiro KOIBUCHI
PY - 2015
DO - 10.1587/transinf.2014EDP7255
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E98-D
IS - 3
JA - IEICE TRANSACTIONS on Information
Y1 - March 2015
AB - Recently network bandwidth becomes a performance concern particularly for collective communication since bisection bandwidths of supercomputers become far less than their full bisection bandwidths. In this context we propose the use of a network coding technique to reduce the number of unicasts and the size of data transferred in latency-sensitive collective communications in supercomputers. Our proposed network coding scheme has a hierarchical multicasting structure with intra-group and inter-group unicasts. Quantitative analysis show that the aggregate path hop counts by our hierarchical network coding decrease as much as 94% when compared to conventional unicast-based multicasts. We validate these results by cycle-accurate network simulations. In 1,024-switch networks, the network reduces the execution time of collective communications as much as 70%. We also show that our hierarchical network coding is beneficial for any packet size.
ER -