|Volume 7, Issue 1 -
Equipment Ugrades Enhance Collaborative Research at CRPC Sites
The computing laboratories at CRPC sites across the country form one of the most advanced collections of high-performance parallel systems in the world. This collection is being expanded throughout 1998 to include a variety of new systems that are serving to enhance existing projects and support new directions in CRPC research.
Earlier this year, the CRPC Facilities Committee approved proposals from the following five CRPC sites to purchase and install cutting-edge systems that will have great impact on the center's education and research efforts.
The CRPC at Caltech focused on obtaining equipment to support large-scale metacomputing demonstrations. A Sun Ultra Sparc Server will replace the existing Sun 690MP Supersparc-based login servers. This provides the primary interactive environment to the site's facilities and serves as a Java development environment incorporating Sun's JavaPlan and Java Workshop development toolsets.
ATM and ethernet routers and switches were acquired to provide a core networking infrastructure capable of handling moderate- to high-bandwidth communication among all major computing and storage resources and to connect them to the very high Bandwidth Networking Service (vBNS). A new tape library consisting of 100 DLT-IV 35 GB tapes provides system backup and specialized archival capacity.
The combination of this equipment, supplemented by Caltech's large parallel computers, HPSS data archive, and ImmersaDesk (a 3D semi-immersive visual reality imaging device purchased largely with CRPC funds), will enable demonstrations such as telecollaboration and visualization with ImmersaDesks at Rice University and Argonne National Laboratory and supporting web tools at Syracuse University.
The CRPC, in partnership with other research groups within Rice, gained approval for the purchase of an ImmersaDesk and OC-12 networking equipment. The OC-12 will connect four parallel computers to the ImmersaDesk via a vBNS switch. The combined purchase permits the parallel facilities at Rice to have a powerful visualization capability and helps create a small prototype of a computational grid. A computational grid uses high-speed networks to connect supercomputers, databases, scientific instruments, and advanced display devices to enable high-performance applications that require unique capabilities not available in a single computer or at a single site.
The system is expected to catalyze significant new research directions for researchers in areas like molecular biology, statistics, and computational and applied mathematics, and serve as a magnet for new funding opportunities in education and research. Another benefit is the system's potential to make computational science and engineering classes dramatically more interesting to students, compared with current classes focused primarily on computation.
The Northeast Parallel Architectures Center (NPAC) at Syracuse obtained a Sun Ultra Model 300, Sun Ultra Enterprise 450 Server, and Ultra Enterprise 2 Model 1300 Server. This cluster of servers is being used for three main purposes: distance learning support, a general information resource for the National HPCC Software Exchange (NHSE) and course curricula, and a Globus Cluster that will be used in work with Argonne National Laboratory to port web-based tools on top of Globus. Globus is an award-winning project that is investigating and developing the technologies required to support the construction of a national-scale computational grid (see Spring/Summer 1998 Parallel Computing Research, "CRPC Honors and Awards").
NPAC researchers are using the new systems for the "International Collaborative Web University," an experimental program to be held this fall that involves universities from China, Europe, and the United States.
Researchers at the University of Tennessee requested an Enterprise 3000 Server with four processors. This powerful server allows the continuation of several projects, most notably the NHSE, beyond its current life. The NHSEis an Internet-accessible resource that facilitates the exchange of software and information among research and computational scientists involved with high-performance computing and communications. University of Tennessee researchers have been closely involved with the project since it was initiated by the CRPC in November 1994.
The Texas Institute for Computational and Applied Mathematics (TICAM) at the University of Texas has assembled a "Beowulf"-class parallel host. It has dual-processor Pentium II nodes with a dual network consisting of a 100 MB network used by the operating system and a 1.2 GB network dedicated to application message passing. This system helps integrate TICAM's extensive base of parallel scientific and engineering applications with advanced parallel visualization research. This vertical integration provides new opportunities for research in interactive steering for large-scale parallel operations, in areas such as subsurface flow and chemistry, computational fluid dynamics, electromagnetic scattering, and data structures.
"All of these new systems have become part of our network of distributed facilities that are shared between sites and with collaborators," says CRPC Facilities Director Paul Messina of Caltech. "They will play a significant role in the CRPC's ongoing efforts to conduct groundbreaking research, promote and improve educational efforts on a global basis, and solve important problems in science and industry using parallel computation."