A team led by Indiana University won this year's Supercomputing Bandwidth Challenge by achieving a peak transfer rate of over 18Gbps out of a possible maximum of 20Gbps.
The goal of the competition was to devise a way to fully utilize one 10Gbps path end-to-end from the Supercomputing 2007 convention in Reno back to the competitors' home academic institutions. This goal is important, say Bandwidth Challenge sponsors SC07, because very few high-bandwidth networks fully utilize their potential speeds.
The IU team created its winning entry using the IU Data Capacitor, a system that the team says is used to store and manipulate massive data sets to achieve a sustained data transfer rate of 16.2Gbps. At its peak, the system nearly doubled the peak transfer rate of its nearest competitors, clocking in at 18.21Gbps. The project utilized an international network path that ran through a variety of research networks, including Internet 2, the pan-European GEANT and Germany's DFN.
To demonstrate its network's high speeds, the IU team ran several different computer applications that required large amounts of bandwidth, including the live acquisition of X-ray crystallography data, the electronic preservation of ancient Sanskrit manuscripts and simulations of high-energy physics reactions.
"This project simultaneously pushed the limits of networking and storage technology while demonstrating a reproducible model for remote data management," says Stephen Simms, IU's Bandwidth Challenge project leader. "Best of all, we did this using a variety of research applications that we support every day at Indiana University."
IU's Bandwidth Challenge Team included partners from the Technische Universitaet Dresden, the Rochester Institute of Technology, Oak Ridge National Laboratory and the Pittsburgh Supercomputing Center.
Last year's Bandwidth Challenge was won by a team from the University of Chicago, which achieved sustained transfer rates of 8Gbps and a peak transfer rate of 9.18Gbps.