We describe a new 512-CPU Beowulf cluster with Teraflop performance dedicated to problems in computational astrophysics. The cluster incorporates a cubic network topology based on inexpensive commodity 24-port gigabit switches and point to point connections through the second gigabit port on each Linux server. This configuration has network performance competitive with more expensive cluster configurations and is scaleable to much larger systems using other network topologies. Networking represents only about 9% of our total system cost of USD561K.ThestandardTop500HPLLinpackbenchmarkratingis1.202Teraflopson512CPUssocomputingcostsbythismeasureare0.47/Megaflop. We also describe 4 different astrophysical applications using complex parallel algorithms for studying large-scale structure formation, galaxy dynamics, magnetohydrodynamic flows onto blackholes and planet formation currently running on the cluster and achieving high parallel performance. The MHD code achieved a sustained speed of 2.2 teraflops in single precision or 44% of the theoretical peak.
@article{arxiv.astro-ph/0305109,
title = {High Performance Commodity Networking in a 512-CPU Teraflop Beowulf Cluster for Computational Astrophysics},
author = {John Dubinski and Robin Humble and Ue-Li Pen and Chris Loken and Peter Martin},
journal= {arXiv preprint arXiv:astro-ph/0305109},
year = {2007}
}
Comments
11 pages, 5 figures, technical paper submitted to Supercomputing 2003 Check http://www.cita.utoronto.ca/~dubinski/sc2003/ for animations