Scheduling many-body short range MD simulations on a cluster of workstations and custom VLSI hardware

J. V. Sumanth, David R. Swanson, Hong Jiang

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Molecular dynamics is a powerful technique used to obtain static or dynamic properties of liquids and solids. The sheer computational intensity of many of these simulations demands more computational power than what any uniprocessor system can provide. Fortunately, these simulations can be parallelized, allowing faster execution times on a cluster of workstations. Of late, custom VLSI chips have been designed to provide an alternative to parallel techniques. The MD-GRAPE 2 is one such solution, offering a peak performance of 64 Gflops. We evaluate the performance and cost-effectiveness of various methods used in sequential and parallel molecular dynamics and the MD-GRAPE 2. We then illustrate how MD simulations involving more complex potential functions can be scheduled on parallel machines and the MD-GRAPE 2 simultaneously.

Original languageEnglish (US)
Pages (from-to)166-175
Number of pages10
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3296
StatePublished - Dec 1 2004

Fingerprint

MD Simulation
Molecular Dynamics Simulation
Molecular Dynamics
Molecular dynamics
Scheduling
Hardware
Complex Potential
Cost-effectiveness
Complex Functions
Parallel Machines
Dynamic Properties
Cost effectiveness
Potential Function
Range of data
Execution Time
Cost-Benefit Analysis
Simulation
Chip
Liquid
Evaluate

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

@article{f4ea90ea661f437b91e886d1125cc09f,
title = "Scheduling many-body short range MD simulations on a cluster of workstations and custom VLSI hardware",
abstract = "Molecular dynamics is a powerful technique used to obtain static or dynamic properties of liquids and solids. The sheer computational intensity of many of these simulations demands more computational power than what any uniprocessor system can provide. Fortunately, these simulations can be parallelized, allowing faster execution times on a cluster of workstations. Of late, custom VLSI chips have been designed to provide an alternative to parallel techniques. The MD-GRAPE 2 is one such solution, offering a peak performance of 64 Gflops. We evaluate the performance and cost-effectiveness of various methods used in sequential and parallel molecular dynamics and the MD-GRAPE 2. We then illustrate how MD simulations involving more complex potential functions can be scheduled on parallel machines and the MD-GRAPE 2 simultaneously.",
author = "Sumanth, {J. V.} and Swanson, {David R.} and Hong Jiang",
year = "2004",
month = "12",
day = "1",
language = "English (US)",
volume = "3296",
pages = "166--175",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Scheduling many-body short range MD simulations on a cluster of workstations and custom VLSI hardware

AU - Sumanth, J. V.

AU - Swanson, David R.

AU - Jiang, Hong

PY - 2004/12/1

Y1 - 2004/12/1

N2 - Molecular dynamics is a powerful technique used to obtain static or dynamic properties of liquids and solids. The sheer computational intensity of many of these simulations demands more computational power than what any uniprocessor system can provide. Fortunately, these simulations can be parallelized, allowing faster execution times on a cluster of workstations. Of late, custom VLSI chips have been designed to provide an alternative to parallel techniques. The MD-GRAPE 2 is one such solution, offering a peak performance of 64 Gflops. We evaluate the performance and cost-effectiveness of various methods used in sequential and parallel molecular dynamics and the MD-GRAPE 2. We then illustrate how MD simulations involving more complex potential functions can be scheduled on parallel machines and the MD-GRAPE 2 simultaneously.

AB - Molecular dynamics is a powerful technique used to obtain static or dynamic properties of liquids and solids. The sheer computational intensity of many of these simulations demands more computational power than what any uniprocessor system can provide. Fortunately, these simulations can be parallelized, allowing faster execution times on a cluster of workstations. Of late, custom VLSI chips have been designed to provide an alternative to parallel techniques. The MD-GRAPE 2 is one such solution, offering a peak performance of 64 Gflops. We evaluate the performance and cost-effectiveness of various methods used in sequential and parallel molecular dynamics and the MD-GRAPE 2. We then illustrate how MD simulations involving more complex potential functions can be scheduled on parallel machines and the MD-GRAPE 2 simultaneously.

UR - http://www.scopus.com/inward/record.url?scp=34547452516&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34547452516&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:34547452516

VL - 3296

SP - 166

EP - 175

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -