Distributed Crossbar Switch Hypermeshes: Efficient Networks for Large-Scale Multicomputers

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

This paper compares the relative performance merits of a regular hypergraph topology, referred to as the Distributed Crossbar Switch Hypermesh (DCSH), and the most common graph topologies, namely the torus and hypercube, taking into account the implementation costs. The comparative analysis considers both uniform and nonuniform traffic patterns. It also includes the effects of switching delays on network performance and uses bit-pipeline transmission to lower the effects of long wires. This study concludes that the DCSH represents a potential candidate as a high-performance multicomputer network as it provides better performance characteristics than the torus and hypercube in most considered cases.

Original languageEnglish
Pages (from-to)999-1012
Number of pages14
JournalJournal of Parallel and Distributed Computing
Volume61
Issue number8
DOIs
Publication statusPublished - Aug 2001

Fingerprint

Multicomputers
Network performance
Hypercube
Switch
Torus
Switches
Topology
Network Performance
Hypergraph
Comparative Analysis
High Performance
Pipelines
Traffic
Wire
Costs
Graph in graph theory

Keywords

  • Multicomputer; interconnection network; hypergraph; hypermesh; graph; torus; hypercube; wormhole routing; message latency; performance comparison

ASJC Scopus subject areas

  • Computer Science Applications
  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

@article{5b5cb81d0a3c4352afe9c72c07a875d7,
title = "Distributed Crossbar Switch Hypermeshes: Efficient Networks for Large-Scale Multicomputers",
abstract = "This paper compares the relative performance merits of a regular hypergraph topology, referred to as the Distributed Crossbar Switch Hypermesh (DCSH), and the most common graph topologies, namely the torus and hypercube, taking into account the implementation costs. The comparative analysis considers both uniform and nonuniform traffic patterns. It also includes the effects of switching delays on network performance and uses bit-pipeline transmission to lower the effects of long wires. This study concludes that the DCSH represents a potential candidate as a high-performance multicomputer network as it provides better performance characteristics than the torus and hypercube in most considered cases.",
keywords = "Multicomputer; interconnection network; hypergraph; hypermesh; graph; torus; hypercube; wormhole routing; message latency; performance comparison",
author = "M. Ould-Khaoua",
year = "2001",
month = "8",
doi = "10.1006/jpdc.2001.1739",
language = "English",
volume = "61",
pages = "999--1012",
journal = "Journal of Parallel and Distributed Computing",
issn = "0743-7315",
publisher = "Academic Press Inc.",
number = "8",

}

TY - JOUR

T1 - Distributed Crossbar Switch Hypermeshes

T2 - Efficient Networks for Large-Scale Multicomputers

AU - Ould-Khaoua, M.

PY - 2001/8

Y1 - 2001/8

N2 - This paper compares the relative performance merits of a regular hypergraph topology, referred to as the Distributed Crossbar Switch Hypermesh (DCSH), and the most common graph topologies, namely the torus and hypercube, taking into account the implementation costs. The comparative analysis considers both uniform and nonuniform traffic patterns. It also includes the effects of switching delays on network performance and uses bit-pipeline transmission to lower the effects of long wires. This study concludes that the DCSH represents a potential candidate as a high-performance multicomputer network as it provides better performance characteristics than the torus and hypercube in most considered cases.

AB - This paper compares the relative performance merits of a regular hypergraph topology, referred to as the Distributed Crossbar Switch Hypermesh (DCSH), and the most common graph topologies, namely the torus and hypercube, taking into account the implementation costs. The comparative analysis considers both uniform and nonuniform traffic patterns. It also includes the effects of switching delays on network performance and uses bit-pipeline transmission to lower the effects of long wires. This study concludes that the DCSH represents a potential candidate as a high-performance multicomputer network as it provides better performance characteristics than the torus and hypercube in most considered cases.

KW - Multicomputer; interconnection network; hypergraph; hypermesh; graph; torus; hypercube; wormhole routing; message latency; performance comparison

UR - http://www.scopus.com/inward/record.url?scp=0008749475&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0008749475&partnerID=8YFLogxK

U2 - 10.1006/jpdc.2001.1739

DO - 10.1006/jpdc.2001.1739

M3 - Article

AN - SCOPUS:0008749475

VL - 61

SP - 999

EP - 1012

JO - Journal of Parallel and Distributed Computing

JF - Journal of Parallel and Distributed Computing

SN - 0743-7315

IS - 8

ER -