GOVERNMENT

High-performance blade system brings immediate productivity to researchers studying key problem in astrophysics: British astrophysicists at the University of Exeter know how trying it can be to simulate the birthing ground of new stars. Using 16 processors of a supercomputer located at another university, the scientists needed an entire year to model a cloud of hydrogen and helium gas so vast it would take a beam of light four years to travel from one end to another. The Exeter scientists found that waiting for results on this resource hobbled their productivity, so to tackle one of the remaining fundamental problems of astrophysics they turned to SGI Altix ICE, a new, high-performance computing (HPC) blade system from SGI. When the 128-core, 16TB SGI Altix ICE 8200 system arrived at the university's campus in June, Exeter realized its wait was over. "Our new SGI Altix ICE system was up and running the afternoon it arrived, and we began running benchmarks within 48 hours," said Matthew Bate, professor of theoretical astrophysics at the University of Exeter. "The system manager began training on it the next morning. When a university can achieve rapid productivity on a powerful system like this, it makes for some very excited researchers." Based on the on-site performance Exeter has seen from the SGI Altix ICE system, the gas cloud simulation that previously kept researchers waiting a year should now take only about six weeks. This represents an eight-fold increase in productivity, even on this complex model. Exeter runs a custom, in-house application for its studies. The compressible fluid dynamics code is the first of its kind in the world to combine magnetohydrodynamics (the study of electrically conducting fluids) with radiation transport, which tracks how photons propagate through the fluid. Both types of studies are necessary to simulate the processes that trigger the creation of stars and planets. Exeter scientists then compare their numerical simulations against actual observations captured by space telescopes to refine their understanding of how turbulence creates dense regions of gas -- regions that collapse under their own gravity to form stars. The SGI Altix ICE system provides Exeter's astrophysics group with an HPC solution powerful enough to keep up with its studies. "These observations and computer simulations help us to better understand the timescales for star formation, how single stars and star clusters originate, the distribution of stellar masses, and the circumstellar discs in which planets form," Bate said. "We have also begun to perform simulations of the formation of gas-giant planets to help explain why extra-solar planets have properties that are so different to the planets in our own solar system." Introduced June 26, SGI Altix ICE is a new line of bladed servers purpose-built to handle true HPC applications while minimizing their demands on the data center's space and power. Its ultra-dense rack architecture delivers up to 40 percent more compute performance per floor tile than competing blades. As Exeter's needs escalate, its new SGI solution will be able to keep up: A single SGI Altix ICE 8200 rack can be powered by as many as 512 Intel Xeon processor cores and deliver 6 TFLOPS of performance. The university's system runs Novell SUSE Linux Enterprise Server 10, the most scalable implementation of Linux available. Like all SGI Altix ICE systems, Exeter's turnkey solution arrived pre-integrated for easy "power up and go" deployment, allowing researchers to focus on their work, not on set-up and administration. "The system software came pre-installed, which helped to make system management and training extremely easy," recalled Bate. "We couldn't help but compare our SGI Altix ICE experience to another, smaller cluster that Exeter purchased from a competitor six months earlier. That system just sat there for several weeks before it was properly up and running. The difference was like night and day." For more information on SGI Altix ICE, go to its Web site.