The past year at Lehigh University has seen great transformation and augmentation to the High Performance Computing Facilities. In 2005, Lehigh’s HPC Steering Committee (HPCSC) approved the purchase of a 60 compute node cluster, called Blaze, where each node has dual AMD Opteron 244 64-bit single core processors and 2 Gigabyte of RAM per node.
In 2006, Lehigh’s HPCSC added the Sun Microsystems based 6 Terabytes of network attached storage for special projects to address the growing need for large data sets becoming increasingly necessary for HPC-based research. Building on this core of hardware, in the past 12 months the HPCSC has upgraded and added extensively to the existing HPC machinery made available to the Lehigh community.
In late 2007, a new Linux cluster, Inferno, a 40-node Beowulf cluster with dual Quad-core Xeon 64-bit processor and 16 Gigabytes of RAM per node, was added to the campus computational grid. An additional 80 Terabytes of disk space was added to the cluster dedicated specifically web search research
In July 2008, a new SGI F1200 32-core (8 quad core Intel Xeon 64-bit processors) shared memory Linux computer, called Altair, with 128 Gigabytes of memory and 3 Terabytes of local disk became the main public, interactive compute device for large computational work for CFD, FEM, and molecular dynamics.
Additionally in July 2008, 20 dedicated workstations were purchased for Earth and Environmental Sciences, Chemistry, Physics and Mathematics research and spread across these departments. Each one of these Dell Precision T7400 workstations has a minimum of a single Intel Xeon quad–core 64-bit processor with 4 GB of RAM, 80GB hard disk, and a 256MB graphics card with a 22-inch monitor.
In August, a new Linux cluster, GCM, a 9-node Beowulf cluster with dual Intel Xeon quad-core 64-bit processors and 16 Gigabytes of RAM per node, with 50 Terabytes of storage with Infiniband node interconnects was purchased specifically to support global climate modeling research. Idle cycles on this cluster will be available for use on Lehigh’s computational grid (described below).
Also this summer, a new donation of a 24-node Egenera machine came to campus. This replaces a previous Egenera donation which was older and no longer supportable. Each computational node on this blade system has 4 Intel Xeon 32-bit hyper-threaded processors with 12 GB of RAM per node. Using this blade based system, Lehigh’s HPC services has put into operation an 18-node Linux login farm called LEAF. Anyone can access this login farm by logging into a head node that uses load balancing to provide the best resources for a user’s needs by directing their login session to the machine with the least amount of current processing load.
All of these resources and more are accessible through a batch dispatching subsystem to access dedicated and idles cycles found across campus. In late 2005, Lehigh’s HPCSC decided to tie together all the University’s compute power through the Condor Project’s batch dispatching subsystem. In addition to all HPC computational hardware running various version of Linux, over 600 Intel Pentium IV/Core Duo campus public space Microsoft Windows and Apple Macintosh computers, which are refreshed (incrementally each year) in a 3-years cycle, became available for HPC computational use while students were not using them.
Undergoing, but which will take 3 more years, is a long term project to upgrade all networking bandwidth to everyone’s desktop on campus to a minimum of 1 Gigabit per second speeds. (Physics research laboratories are scheduled for upgrade by first quarter of calendar year 2009.) Users who need more computational resources can go off-campus to national supercomputing facilities, such as the Teragrid, via our 100Mbs bandwidth connection to Internet2.
Last Modified: October 30, 2008