High Performance Computing gets much-needed upgrade
The UA is upgrading its high performance computer (HPC) systems to enable researchers to accomplish more in a shorter period of time.
UA Supercomputer Replacement Project includes two major HPC systems. The first system, a 568 CPU SGI Altix 4700 'shared memory' system was installed in March 2007. The second system — a cluster computer — is being installed and tested in February 2008 and will be made available for researchers in March. The new systems provide a significant improvement of capacity and capabilities and require less space and significantly reduced air conditioning due to the water cooling options installed on the systems.
The University of Arizona has selected SGI® and JRTI® to supply its new HPC cluster computing system. The system selected is a 1392 CPU SGI Altix ICE system (Xeon-Harpertown, 15.76 TFLOPS.) This system will be added to the existing HPC resources at UA which include an SGI 626-core Altix 4700 (3.6 TFLOPS) system with two SGI RC100 dual-core Xilinx FPGA blades.
Collaboration for Success
This most recent purchase is the result of a collaboration led by UA University Information Technology Services (UITS), Learning Technology Services (LTS) Research Computing partnering with eight faculty research groups from across campus. The faculty members are Vicki Chandler (BIO5), Mike Sanderson (Ecology & Evolutionary Biology), Howard Ochman (Biochemistry, Ecology & Evolutionary Biology), Steven Rounsley (BIO5), Romeel Davé (Astronomy), Mike Hammer (Arizona Research Labs), Cari Soderlund (Computer Science, Plant Science) and Joanna Masel (Ecology and Evolutionary Biology). Approximately half of the new cluster will allocated at high priority to the partner research groups in proportion to the funding that they provided.
"HPC systems takes us to a new level of computational ability and enables significant advances in our field," Davé said. Davé is using HPC resources to model the formation of galaxies in large-scale simulations of the universe from the big bang to the present. See examples of this work.
"HPC being so wonderful makes my life easier. My research will really be able to speed up, opening up new questions in systems biology that previously could not be addressed," said Masel, assistant professor in ecology and evolutionary biology.
The HPC systems are operated by UITS’s Infrastructure Services, Computing Infrastructure group and are available to all UA research groups to support a wide variety of research across campus.
The HPC resources are designed to be expandable so that faculty research groups can continue to add to the cluster in the future. The Altix 4700 is currently online and available for all users.
Green Machines
The previous system had an air conditioning cooling requirement of 29 tons. The new SGI Altix computer systems use the University chilled water system for cooling to reduce the requirements for AC cooling. Craig Cook, UITS Computer Operations Manager, has reduced the total machine room AC cooling requirements from 50 tons to 30 tons. Gordon Bush, from UA Utilities, has indicated that the chilled water system is approximately twice as efficient as air conditioning for room cooling. The reduction in AC cooling corresponds to an overall 20 percent savings in energy costs for computing systems cooling. At the same time, HPC computing capacity is increasing from 0.6 TeraFLOPS (TFLOPS) to a combined capacity of 19.4 TFLOPS, over 32 times the previous capacity.
Detailed information on the HPC resources and support is available.
Learning Technology Services (LTS) is a campus resource for The University of Arizona, providing support for faculty, all instructional personnel and researchers to integrate technology into academic activities. The LTS is a unit of the Frontline Services division of University Information Technology Services (UITS) and was formerly known as the Learning Technologies Center (LTC).
RSS Feed: