APPLICATIONS
USC's Supercomputing Cluster Boasts Nearly a 50% Increase in Compute Speed
Sun Microsystems today announced that the University of Southern California (USC) Center for High-Performance Computing and Communications (HPCC), a worldwide leader among research computing centers, has added Sun Microsystems' leading dual-core Sun Fire V20z x64 (x86, 64-bit) servers powered with the AMD Opteron processors to its powerful supercomputer cluster. USC purchased 360 dual-core Sun Fire V20z servers for the HPCC supercomputer cluster, selecting them for their advanced cooling capabilities, smaller form factor, energy efficiency and superior price/performance. Driven by a steady desire to strengthen the performance and capability of its supercomputing server cluster, which is currently ranked as the 24th-fastest supercomputer in the world on the Top500 list, the recent addition of the Sun Fire V20z servers increased USC's HPCC system total number of compute nodes from 1,716 to 1,830.
Preliminary testing performed at the Sun Solution Center for High Performance Computing (HPC) in Hillsboro, Ore., clearly demonstrated that Sun's servers can deliver near 85 percent of the peak floating point performance on the Linpack Benchmark, which is a highly parallel computing test designed to measure how fast a computer system can solve linear equations, a common task in engineering and scientific applications. The HPCC recently ran this benchmark using Sun's industry-standard x64 servers and achieved a result of 10.75 TeraFLOPS (TFLOPS) or 10.75 trillion calculations per second. This resulted in the fastest computation USC has ever processed, delivering a nearly 50 percent increase over its previous Linpack Benchmark results.
"We look to Sun as one of our key partners when augmenting our HPCC environment. When we wanted to boost the speed and power of our supercomputer, we decided to use the dual-core Sun Fire V20z server," said Mike Pearce, USC's deputy CIO. "The dual-core Sun Fire V20z servers helped us improve our benchmark results and compete head-to-head with more heavily funded government-supported supercomputing projects around the world without increasing the system footprint and power consumption requirements, which can
be a major expense."
USC's HPCC supports large-scale supercomputing projects for more than 120 research groups at the university in a variety of disciplines, ranging from genomics and geophysics to materials science and natural language translation.
USC's HPCC has quickly risen to become a leader among supercomputing centers and makes steady IT investments to achieve its current status among the world's top supercomputer sites.
"Our HPC and grid computing solutions help organizations like USC to maximize productivity, reduce cost and complexity, improve ROI and speed time-to-market. Sun offers flexibility and choice for HPC with a complete portfolio of technologies and partnerships for a successful implementation," said Marc Hamilton, director of technology for Sun's Global Education and Research Group. "Our continued innovation around Sun's x64 servers has resulted in record-breaking performance for USC's HPCC server cluster, allowing them to demolish previous records by adding a critical boost of extra power which complemented its supercomputer's existing server nodes."
Following its investment in Sun Fire V20z servers, USC recently implemented an enterprise-class Sun Fire x64 server in its HPCC. USC selected the new Sun Fire X4200 server powered with the AMD Opteron processor to evaluate in its center based on its performance capabilities and ease-of-management. USC anticipates that the new Sun Fire x64 servers will help its supercomputer run faster with less maintenance.
"Early use of the Sun Fire X4200 server has proved successful," added Pearce. "The Sun Fire x64 server enables us to look at scaling our supercomputer even more in the future based on its advanced systems management capabilities, enhanced power and new memory system."
More information on USC's Center for High-Performance Computing and Communications can be found at: its Web site .