Supercomputers

The Supercomputing Institute has continued the strong tradition of providing University of Minnesota researchers with leading-edge, high-performance computing technologies and diversified programs that complement these technologies. In addition, the Institute has developed a strong program of user support, including tutorials and applications support across the physical, biological, mathematical, and computer sciences, engineering, and other disciplines that use high-performance computing, informatics, and data mining.

In April 2000, the WinterHawk nodes were upgraded to WinterHawk+ technology. The Institute’s IBM SP supercomputer currently consists of a 370-processor supercomputer with 663 GB of memory. The nodes available for computation are:

• 74 4-processor WinterHawk+ nodes with 375 MHz Power3 processors; 16 nodes have 8 GB of memory each; the remaining nodes each have 4 GB of memory.
• 12 4-processor NightHawk nodes, each with 222 MHz Power3 processors and 16 GB of memory. Four of the WinterHawk+ nodes are available for interactive use, and the remaining WinterHawk+ nodes and all NightHawk nodes are for computation through the queuing system. There are also four WinterHawk+ nodes used as file servers.

In addition, the Institute has a 16-processor Silvernode IBM SP with 12 GB of memory. The Silvernode SP utilizes the 332 MHz PowerPC 604e processor and has 4 TB of disk space. The IBM supercomputer resources are available in coordination with the IBM Shared University Research (SUR) partnership.

In March 2002, the Supercomputing Institute purchased an IBM pSeries 690 (Regatta). This system was upgraded in August 2002 to two Regattas. The system consists of two compute nodes with 1.3 GHz Power4 processors, file servers, and an interactive node. One of the compute nodes has 32 processors sharing 64 GB of memory, while the second node has 24 processors sharing 24 GB of memory. The interactive node is a WinterHawk+ node. Four 4-processor NightHawk nodes are used as file servers for 7 TB of disk.

The Institute’s LINUX Cluster consists of 82 2-processor Netfinity nodes from IBM. Of these 82 nodes, 76 are available for computation, five are file servers, and one node is for interactive use. Seventeen of these nodes have two 733 MHx Intel Pentium III processors sharing 1 GB of memory. The remaining nodes have two 1.26 GHz Pentium III processors sharing 3.25 GB of memory. The nodes are connected together using Fast Ethernet and Myrinet. A cluster file system provides 2.5 TB of disk that can be used on all nodes.

The IBM systems will be upgraded in July 2003 and further upgraded in Fall 2003.

In May 2001, the Supercomputing Institute installed two 96-processor SGI Origin 3800s. Each SGI Origin 3800 has 144 GB of memory. These supercomputers utilize 500 MHz R14000 processors. The Institute also offers access to a 12-processor SGI Origin 2000 with 22 GB of memory. This Origin system will be replaced in June 2003 by a new SGI Altix system.

In 2003, the Institute installed a 32-processor Intel Xeon 2.0 GHz ES7000 Orion 230 supercomputer with 64 GB of memory, Windows 2000 Datacenter operating system software, and a 32-processor Microsoft SQL Server license.

These supercomputers offer the Supercomputing Institute’s researchers access to state-of-the-art, high-performance computing technology. The Supercomputing Institute joined the Digital Technology Center in 2001, and in 2002 the supercomputing resources moved to new facilities on the East Bank of the Minneapolis Campus. In addition, the Supercomputing Institute is continuing its commitment to a diversified array of computing laboratories, collaborations, and programs. These include the Basic Sciences Computing Laboratory, the Computational Genetics Laboratory, the Digital Technology Computational Biology Laboratory, the Scientific Development and Visualization Laboratory, the Medicinal Chemistry/Supercomputing Institute Visualization–Workstation Laboratory, the Laboratory for Large-Scale Data Analysis, the Laboratory for Computational Science and Engineering, and interdisciplinary Ph.D. programs in Scientific Computing and Computational Neuroscience. The major supercomputing resource program and long-term planning at the Institute are guided by the Institute’s Planning Committee.

Planning Committee

Donald G. Truhlar, Supercomputing Institute Director, Chemistry, Chemical Physics, and Scientific Computation,
Graham V. Candler, Aerospace Engineering and Mechanics and Scientific Computation
James R. Chelikowsky, Chemical Engineering and Materials Science and Scientific Computation
Jiali Gao, Chemistry and Scientific Computation
Yiannis Kaznessis, Chemical Engineering and Materials Science
Vipin Kumar, Computer Science and Engineering and Scientific Computation
David J. Lilja, Electrical and Computer Engineering and Scientific Computation
Alon V. McCormick, Chemical Engineering and Materials Science
Douglas H. Ohlendorf, Biochemistry, Molecular Biology, and Biophysics
Yousef Saad, Computer Science and Engineering and Scientific Computation
George L. Wilcox, Neuroscience and Scientific Computation