FOR IMMEDIATE RELEASE CONTACT: February 6, 1998 Michael Schneider Pittsburgh Supercomputing Center 412-268-4960 firstname.lastname@example.org
Pittsburgh, PA With CRAY T3E systems on both sides of the Atlantic working in tandem, scientists in Stuttgart and Pittsburgh conducted what they believe is the largest simulation of granular gas ever carried out. The simulation, conducted in November at Supercomputing '97 in San Jose, included 1.75 billion particles, 500 million more than the largest previously known granular particle simulation.
The simulation used the Transatlantic Metacomputing Testbed established since June 1997 between the High Performance Computing Center at Stuttgart University (HLRS) and the Pittsburgh Supercomputing Center (PSC). This link, the first successful prototype for transatlantic metacomputing, employs high-performance research networks in the United States, Canada and Germany to couple a 512-processor T3E at PSC with another at HLRS.
By linking the two systems, creating a virtual T3E of 1,024 processors, the researchers obtained enough computing power to increase the size of the simulation from 1.2 to 1.75 billion particles. The researchers' objective was to demonstrate the potential of metacomputing.
"Metacomputing is being developed as a tool to solve large problems," says Matthias Mueller, a scientist at the University of Stuttgart's Institute of Computer Applications, which developed the granular particle simulation software. "So far it has been a technical issue to develop that tool. Demonstrations such as ours show what has to be done to make reasonable real simulations possible."
Simulations of such huge numbers of particles are needed to understand complex granular processes, says Mueller, many of which are important in industry and commerce. Examples include crack propagation in concrete and storage and shipment of foodstuffs such as grains, flour and sugar. Since early in this century, researchers have studied these materials to improve industrial processes, but fundamental questions remain. Scientists still lack a detailed theoretical understanding, for instance, of how pipes carrying granular materials become clogged or why grains of different size separate and gather in band-like patterns.
Communications software developed at HLRS distributes the granular particles of the calculation among the 1,024 processors of the two T3Es and allows the two systems to communicate with each other. The coupled system, note the researchers, operated with only 75 millisecond latency, compared to unavoidable latency of 20 msec due to the speed of light.
Since starting their transatlantic metacomputing collaboration, PSC and HLRS have improved latency and asynchronous communication. The main problem now, says Sergiu Sanielevici, manager of parallel applications research at PSC, appears to be the bandwidth of the connection. In future work, the researchers expect to improve the execution speed of metacomputing applications by better latency hiding and optimized communications software.
"The good news," says Sanielevici, "is that this demonstration moves us a step closer to a global information infrastructure. We're showing that we have in place the capability to collaborate effectively with the European community."
For more information, see http://www.psc.edu/publicinfo/news/1997/rus_meta.html or http://www.hlrs.de/news/events/sc97.html
The Pittsburgh Supercomputing Center is a joint effort of Carnegie Mellon University and the University of Pittsburgh together with Westinghouse Electric Corp. It was established in 1986 and is supported by several federal agencies, the Commonwealth of Pennsylvania and private industry.