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

Authors:
J. V. Sumanth;David R. Swanson;and Hong Jiang
Affiliations:
Department of Computer Science and Engineering, University of Nebraska-Lincoln, Lincoln, NE;Department of Computer Science and Engineering, University of Nebraska-Lincoln, Lincoln, NE;Department of Computer Science and Engineering, University of Nebraska-Lincoln, Lincoln, NE
Venue:
HiPC'04 Proceedings of the 11th international conference on High Performance Computing
Year:
2004

Citing 3
Cited 2

Solving problems on concurrent processors

Solving problems on concurrent processors
Computer simulation of liquids

Computer simulation of liquids
Performance and cost effectiveness of a cluster of workstations and MD-GRAPE 2 for MD simulations

ISPDC'03 Proceedings of the Second international conference on Parallel and distributed computing

A symmetric transformation for 3-body potential molecular dynamics using force-decomposition in a heterogeneous distributed environment

Proceedings of the 21st annual international conference on Supercomputing
A novel force matrix transformation with optimal load-balance for 3-body potential based parallel molecular dynamics using atom-decomposition in a heterogeneous cluster environment

HiPC'07 Proceedings of the 14th international conference on High performance computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Molecular dynamics is a powerful technique used toobtain static or dynamic properties of liquids and solids The sheercomputational 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.