H-BSP

a hierarchical BSP computation model

Hojung Cha, Dongho Lee

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

This paper presents a new parallel computing model, called H-BSP, which adds a hierarchical concept to the BSP (Bulk Synchronous Parallel) computing model. An H-BSP program consists of a number of BSP groups which are dynamically created at run time and executed in a hierarchical fashion. H-BSP allows algorithm designers to develop more efficient algorithms by utilizing processor locality in the program. Based on the distributed memory model, H-BSP provides a group-based programming paradigm and supports Divide & Conquer algorithms efficiently. This paper describes the structure of the H-BSP model, complexity analysis and some examples of H-BSP algorithm. Also presented is the performance characteristics of H-BSP algorithms based on the simulation analysis. Simulation results show that H-BSP takes advantages of processor locality and performs well in low bandwidth networks or in a constant-valence architecture such as 2-dimensional mesh. It is also proved that H-BSP can predict algorithm performance better than BSP, due to its locality-preserving nature.

Original languageEnglish
Pages (from-to)179-200
Number of pages22
JournalJournal of Supercomputing
Volume18
Issue number2
DOIs
Publication statusPublished - 2001 Feb 1

Fingerprint

Parallel Computation
Parallel algorithms
Parallel processing systems
Locality
Parallel Algorithms
Parallel Computing
Model
Model Complexity
Bandwidth
Memory Model
Complexity Analysis
Data storage equipment
Parallel Programs
Distributed Memory
Simulation Analysis
Divides
Efficient Algorithms
Programming
Paradigm
Mesh

All Science Journal Classification (ASJC) codes

  • Software
  • Theoretical Computer Science
  • Information Systems
  • Hardware and Architecture

Cite this

Cha, Hojung ; Lee, Dongho. / H-BSP : a hierarchical BSP computation model. In: Journal of Supercomputing. 2001 ; Vol. 18, No. 2. pp. 179-200.
@article{5f1e74cda47f4591baa6a66fd9666ca3,
title = "H-BSP: a hierarchical BSP computation model",
abstract = "This paper presents a new parallel computing model, called H-BSP, which adds a hierarchical concept to the BSP (Bulk Synchronous Parallel) computing model. An H-BSP program consists of a number of BSP groups which are dynamically created at run time and executed in a hierarchical fashion. H-BSP allows algorithm designers to develop more efficient algorithms by utilizing processor locality in the program. Based on the distributed memory model, H-BSP provides a group-based programming paradigm and supports Divide & Conquer algorithms efficiently. This paper describes the structure of the H-BSP model, complexity analysis and some examples of H-BSP algorithm. Also presented is the performance characteristics of H-BSP algorithms based on the simulation analysis. Simulation results show that H-BSP takes advantages of processor locality and performs well in low bandwidth networks or in a constant-valence architecture such as 2-dimensional mesh. It is also proved that H-BSP can predict algorithm performance better than BSP, due to its locality-preserving nature.",
author = "Hojung Cha and Dongho Lee",
year = "2001",
month = "2",
day = "1",
doi = "10.1023/A:1008113017444",
language = "English",
volume = "18",
pages = "179--200",
journal = "Journal of Supercomputing",
issn = "0920-8542",
publisher = "Springer Netherlands",
number = "2",

}

H-BSP : a hierarchical BSP computation model. / Cha, Hojung; Lee, Dongho.

In: Journal of Supercomputing, Vol. 18, No. 2, 01.02.2001, p. 179-200.

Research output: Contribution to journalArticle

TY - JOUR

T1 - H-BSP

T2 - a hierarchical BSP computation model

AU - Cha, Hojung

AU - Lee, Dongho

PY - 2001/2/1

Y1 - 2001/2/1

N2 - This paper presents a new parallel computing model, called H-BSP, which adds a hierarchical concept to the BSP (Bulk Synchronous Parallel) computing model. An H-BSP program consists of a number of BSP groups which are dynamically created at run time and executed in a hierarchical fashion. H-BSP allows algorithm designers to develop more efficient algorithms by utilizing processor locality in the program. Based on the distributed memory model, H-BSP provides a group-based programming paradigm and supports Divide & Conquer algorithms efficiently. This paper describes the structure of the H-BSP model, complexity analysis and some examples of H-BSP algorithm. Also presented is the performance characteristics of H-BSP algorithms based on the simulation analysis. Simulation results show that H-BSP takes advantages of processor locality and performs well in low bandwidth networks or in a constant-valence architecture such as 2-dimensional mesh. It is also proved that H-BSP can predict algorithm performance better than BSP, due to its locality-preserving nature.

AB - This paper presents a new parallel computing model, called H-BSP, which adds a hierarchical concept to the BSP (Bulk Synchronous Parallel) computing model. An H-BSP program consists of a number of BSP groups which are dynamically created at run time and executed in a hierarchical fashion. H-BSP allows algorithm designers to develop more efficient algorithms by utilizing processor locality in the program. Based on the distributed memory model, H-BSP provides a group-based programming paradigm and supports Divide & Conquer algorithms efficiently. This paper describes the structure of the H-BSP model, complexity analysis and some examples of H-BSP algorithm. Also presented is the performance characteristics of H-BSP algorithms based on the simulation analysis. Simulation results show that H-BSP takes advantages of processor locality and performs well in low bandwidth networks or in a constant-valence architecture such as 2-dimensional mesh. It is also proved that H-BSP can predict algorithm performance better than BSP, due to its locality-preserving nature.

UR - http://www.scopus.com/inward/record.url?scp=0035250519&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0035250519&partnerID=8YFLogxK

U2 - 10.1023/A:1008113017444

DO - 10.1023/A:1008113017444

M3 - Article

VL - 18

SP - 179

EP - 200

JO - Journal of Supercomputing

JF - Journal of Supercomputing

SN - 0920-8542

IS - 2

ER -