This Article Full Text (PDF) References Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Add to Saved Citations Download to citation manager Request Permissions Request Reprints Add to My Marked Citations Citing Articles Citing Articles via Google Scholar Citing Articles via Scopus Google Scholar Articles by Rycerz, K. Articles by Sloot, P. M. A. Search for Related Content Social Bookmarking                         What's this?

Regular Paper: Interactive N-Body Simulations On the Grid: HLA Versus MPI

INSTITUTE OF COMPUTER SCIENCE, AGH, AL. MICKIEWICZA 30, 30-059 KRAKÓW, POLAND, KZAJAC{at}AGH.EDU.PL

Alfredo Tirado-Ramos

FACULTY OF SCIENCE, SECTION COMPUTATIONAL SCIENCE, UNIVERSITY OF AMSTERDAM, KRUISLAAN 403, 1098 SJ AMSTERDAM, THE NETHERLANDS

Alessia Gualandris

FACULTY OF SCIENCE, SECTION COMPUTATIONAL SCIENCE, UNIVERSITY OF AMSTERDAM, KRUISLAAN 403, 1098 SJ AMSTERDAM, THE NETHERLANDS, ASTRONOMICAL INSTITUTE "ANTON PANNEKOEK", AMSTERDAM,THE NETHERLANDS

Simon F. Portegies Zwart

FACULTY OF SCIENCE, SECTION COMPUTATIONAL SCIENCE, UNIVERSITY OF AMSTERDAM, KRUISLAAN 403, 1098 SJ AMSTERDAM, THE NETHERLANDS, ASTRONOMICAL INSTITUTE "ANTON PANNEKOEK", AMSTERDAM,THE NETHERLANDS

Marian Bubak

INSTITUTE OF COMPUTER SCIENCE, AGH, AL. MICKIEWICZA 30, 30-059 KRAKÓW, POLAND, FACULTY OF SCIENCE, SECTION COMPUTATIONAL SCIENCE, UNIVERSITY OF AMSTERDAM, KRUISLAAN 403, 1098 SJ AMSTERDAM, THE NETHERLANDS

Peter M. A. Sloot

FACULTY OF SCIENCE, SECTION COMPUTATIONAL SCIENCE, UNIVERSITY OF AMSTERDAM, KRUISLAAN 403, 1098 SJ AMSTERDAM, THE NETHERLANDS

The grid paradigm for distributed computation provides an interesting framework for the simulation of dense stellar systems, which remains a great challenge for astrophysicists. In this work we apply a distributed simulation model to the astrophysical N-body problem: the High Level Architecture (HLA), using distributed federations on top of grid infrastructures for the communication between simulation and visualization components. To achieve this goal, we use a grid HLA Management system (G-HLAM) for HLA-based simulations running on the grid. We compare this setup with a direct parallelization of the N-body code using MPI on the grid. Our aim is to provide scientists with an interactive environment where, in contrast to traditional MPI-based parallel processing, seamless simulation process migration and rollbacks can be invoked on the fly to improve the overall computational performance.

We found that MPI and HLA are complementary rather than competing technologies, as they can be used simultaneously to improve the performance of complex simulations on different fronts. On one hand, if the communication to computation ratio is small enough, the MPI parallelization proves to be a relatively easy and efficient method for taking advantage of grid technology. On the other hand, HLA provides advanced mechanisms to synchronize simulation and visualization components located on different grid nodes and can be used to add steering mechanisms to improve interaction. Finally, we found that HLA allows migration of running simulations to be easily managed, reducing overall communication times between simulation and the user, effectively improving the overall system performance.

Key Words: interactive simulation • HLA • grid • MPI • N-body

International Journal of High Performance Computing Applications, Vol. 21, No. 2, 210-221 (2007)
DOI: 10.1177/1094342007074874


CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    Twitter    What's this?