New High Performance Computing Cluster Arrives

The most recent addition to the array of high performance computing hardware on campus arrived in the first week of October. Library and Technology Services (LTS) has purchased a 40-node Beowulf computer cluster bringing an amazing amount of new computational power to the University.

Scientific Computing Team Leader Gale Fritsche is dwarfed by one half of "Inferno," the new HPC Beowulf cluster.

Each of the 40 computer nodes consists of two quad-core Intel Xeon 5320 processors with 16 Gigabytes of RAM per node. The new cluster is called Inferno. LTS worked closely with the High Performance Computing (HPC) Steering Committee, led by Associate Professor Ted Ralphs in Industrial and Systems Engineering, to recommend a system and a vendor for this purchase to LTS Vice Provost Bruce Taggart.

 This new cluster is part of the computational grid LTS has configured using the Condor project software, from the University of Wisconsin, as a batch processing subsystem. Researchers on campus who need large amounts of computational power can take advantage of these HPC resources as well as idle computers in public spaces on this grid.

For example, Associate Professor Bruce Dodson in the Mathematics Department was able to amass over 466 years of computer time in the past nine months to further his research on prime numbers.

Through a cooperative effort between LTS and the Computer Science and Engineering Department, Assistant Professor Brian Davison has contributed funds to this cluster purchase for an additional 80 Terabytes of disk space to be used in his web search research. Dr. Davison is utilizing the cluster to advance knowledge in such areas as “cloud computing” which hopes to make the parallelization of basic operations that speedup web searches, such as alphanumeric string searches, transparent to developers of new tools.

This new computer cluster technology will provide researchers on campus access to hardware faster and more powerful than the previous cluster, Blaze, purchased in 2005. Each node in the cluster Inferno has 8 times more RAM and twice as many computer cores as Blaze, providing up from two to five times the computational throughput. For more information on how to take advantage of Lehigh’s HPC resources, please visit here.

-- Brandon Leeds,
   Senior Computer Consultant

Article posted November 2007
 

Return to Newsletter