Performance of a parallel astrophysical n-body solver on pan-european computational grids

  • Authors:

  • Alfredo Tirado-Ramos;Alessia Gualandris;Simon Portegies Zwart

  • Affiliations:

  • Faculty of Science, Section Computational Science, University of Amsterdam, Amsterdam, The Netherlands;Faculty of Science, Section Computational Science, University of Amsterdam, Amsterdam, The Netherlands;Faculty of Science, Section Computational Science, University of Amsterdam, Amsterdam, The Netherlands

  • Venue:
  • EGC'05 Proceedings of the 2005 European conference on Advances in Grid Computing
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present performance results obtained by running a direct gravitational N-body code for astrophysical simulations across the Dutch DAS-2 and the pan-European CrossGrid computational grids. We find that the performance on large grids improves as the size of the N-body system increases because the computation to communication ratio becomes higher and a better load balance can be achieved. Communication among nodes residing in different locations across Europe becomes more evident as the number of locations increases. Nevertheless, contrary to our expectations, we find that the performance decreases only by about a factor three for a large simulation. We conclude that highly distributed computational Grid infrastructures can be used efficiently for simulating large gravitational N-body systems.