Intended as a prototype for international high-performance networking, the project couples Pittsburgh's 512-processor CRAY T3E with another 512-processor T3E at the High Performance Computing Center in Stuttgart (RUS). Linking two or more supercomputers at different locations in this manner, for work on the same computing task, is known as "metacomputing." The Pittsburgh-Stuttgart link creates a virtual system of 1024 processors with a theoretical peak performance of 675 billion calculations a second.
"There are many large-scale research problems," said Ralph Roskies and Michael Levine, co-scientific directors of the Pittsburgh Supercomputing Center, "such as modeling global climate, simulating automobile crashes with people inside and the design of new drug therapies, that demand the expanded capability potentially available through metacomputing. Our center has been at the forefront of this work for some time, and our collaboration with Stuttgart is an important step toward proving the viability of this concept."
The PSC-RUS team is using the coupled system to run flow simulation software, called URANUS, that predicts aerodynamic forces and high temperatures acting on space vehicles when they reenter Earth's atmosphere. To date, the researchers have used up to 64 processors simultaneously on both T3E systems.
PSC computational scientists Bruce Loftis and Raghurama Reddy collaborated with an RUS group, led by Michael Resch, for about a year to develop and test software that allows the two T3Es to work together on the same problem. "This is a highly complex programming task," said Loftis, "and what we've accomplished is a starting point for further work."
The project relies on a series of research networks to create a high-speed transatlantic link between the two centers. Such networks, established during the past few years, allow information to move up to 100 times faster than on the Internet. The vBNS, for example, which connects U.S. supercomputing centers, currently transmits at speeds up to 622 million bits per second, fast enough to transfer the complete Encyclopedia Britannica in less than 10 seconds.
The Pittsburgh to Stuttgart connection goes from PSC via vBNS to STAR TAP, a National Science Foundation project that provides a U.S. interconnection point for high-speed networking with research institutions worldwide. Two Canadian networks provide the key intermediary links between the U.S. and Germany. STAR TAP connects to the Canadian Network for the Advancement of Research, Industry and Education (CANARIE), which in turn connects to Teleglobe, a transatlantic link, that connects to Deutsche Telekom AG (DTAG), the German research network, that connects to RUS. "Today's successful T3E-to-T3E connection culminates months of hard work at PSC, RUS, MCI, STAR TAP, CANARIE, Teleglobe Canada and DTAG," said PSC parallel applications manager Sergiu Sanielevici.
"The key to metacomputing, now and in the future," said Wendy Huntoon, PSC networking manager, "is very high bandwidth networking such as vBNS. This project shows that we now have in place the networking capability to collaborate effectively with the European community, a long awaited step."
In the next few months, the Pittsburgh-Stuttgart team will expand their metacomputing project to include a third supercomputer, an Intel Paragon at Sandia National Laboratory.
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.