High Performance Computing

Notre Dame Adds 6,000-Core HPC Cluster to Center for Research Computing

• By Dian Schaffhauser
• 04/26/10

The University of Notre Dame has gone public with its installation of a high performance cluster for its scientific research computing work. After an extensive evaluation in 2009 that assessed systems from Sun Microsystems, HP, IBM, Dell, SGI, and Penguin Computing, the university chose HP Model DL165 G6 servers to add a nearly 6,000 core cluster to its Center for Research Computing.

Each node contains dual AMD six-core Opteron processors, providing a total of 12 cores per node. These HP systems offer 1 GB of RAM per core, for a total of 12 GB, and a 160 GB SATA disk with about 100 GB of that available for "scratch" space.

"Innovative research is one of the pillars of research at Notre Dame," said Robert Bernhard, vice president for research at the university. "This new cluster is a significant step forward in expanding our computational capabilities. We're anticipating using the cluster to explore challenging and important questions across many disciplines at Notre Dame."

The university said 50 groups are already using the CRC cluster, for research projects such as modeling blood clotting processes to advance the understanding and treatment of heart disease, strokes, and hemophilia; developing a molecular-level understanding of chemical reactivity at surfaces; and creating computational models of ocean hydrodynamics that can be applied to problems such as coastal flooding.

"The new CRC cluster has already allowed some researchers to more than triple their productivity," said Jarek Nabrzyski, director of the computing center. "This new cluster was also surprisingly easy to integrate into our existing systems, and the excellent high-performance-to-price ratio allows us to maximize new resource allocations for our high capacity user requirements."

Launching the new CRC cluster is part of a $1.8 million research computing investment. Future projects will include bringing in additional clusters, cluster nodes, an SMP system, and additional high-performance storage. The most recent integration was handled by Matrix Integration, a technology solutions provider based in Jasper, IN.