SCIENCE
Lawrence Livermore National Labs Selects Voltaire & Appro
SPECIAL COVERAGE FROM ISC2006 - Voltaire 20 Gigabit/Second InfiniBand Solutions To Power 16,128-Processor Clustered System Supporting Large-Scale ComputationalResearch -- Voltaire, the worldwide leader in grid backbone solutions, today announced that Lawrence Livermore National Laboratories (LLNL) has selected Voltaire Grid Backbone solutions as part of an Appro (SC Online Product of the year award winner * Please see our article at http://www.supercomputingonline.com/article.php?sid=10854) clustered supercomputer for the Peloton Linux Cluster Project. The 16,128 processor supercomputer will leverage Voltaire's multi-service director-class switches with double data rate (DDR), 20 gigabit per second InfiniBand to provide industry-leading performance for LLNL's complex computational research. The supercomputer will consist of three Appro 1U Quad XtremeServer Linux clusters with a total of 2,016 quad socket nodes and 16,128 cores based on the latest AMD dual-core processors. Multiple Voltaire Grid Director ISR 9288 multi-service switches and multiple Grid Switch ISR 9024 edge switches will form the high performance, DDR InfiniBand fabric. Voltaire DDR InfiniBand switches and software deliver double the performance of existing 10 Gigabits/second interconnect offerings and provide unparalleled bandwidth, performance and scalability. 
"We are extremely pleased to work with Appro to deliver this world-class system to LLNL,"said Peter Waxman, vice president sales, North America at Voltaire. "Voltaire's DDR InfiniBand solutions enable LLNL to reap the additional processing power provided by dual-core machines to be able to run complex calculations and scientific simulations faster and more efficiently." Two of the three Linux clusters will be used in the Laboratory's unclassified environment as a multi-programmatic and institutional (M&IC) resource. The other Linux cluster will be used in the classified environment to solve complex computational problems related to the Laboratory's National Nuclear Security Administration's (NNSA) program to ensure the safety, security and reliability of the nation's nuclear deterrent without testing. This program is known as the Stockpile Stewardship. Identical scalable units based on twelve, Voltaire 24-port Grid Switch(TM) ISR 9024 DDR switches with 144 quad socket nodes and 1,152 cores each will be grouped together with multiple Voltaire 288-port Grid Director(TM) ISR 9288 DDR switches to form the three Linux clusters. Multiple organizations and programs within the Laboratory will share these clusters for large, medium and small scale scientific simulations. "Peloton scalable unit deployments are the next stage in Linux cluster innovation that will enable Livermore's M&IC next generation scientific simulation capability in support of our national security missions," said Mark Seager, ICCD ADH for Advanced Technology. "With the Peloton project, the Laboratory will deploy InfiniBand interconnect into large scale production resources for the first time. Voltaire's DDR solution will significantly increase the bandwidth and messaging rate required to meet the ever-increasing demand for the large scale simulations indispensable to advancing all scientific disciplines. This DDR network is well-balanced with Quad socket, dual core Opteron nodes for our very demanding mission-driven scientific simulations and allows us to build very large Linux clusters with improved MTBF characteristics." "We are very pleased to work with Voltaire's 20 Gbps InfiniBand interconnect to power Appro XtremeServer Supercomputing clusters to meet the staggering demand for 50-100 teraFLOP/s scale scientific simulations," said Daniel Kim, CEO of Appro. "Appro delivers a totally managed HPC solution that can incorporate any of our full range of high-performance server platforms and technologies building a cost-effective, high-bandwidth clusters today on an infrastructure that is flexible and easy to scale."
"We are extremely pleased to work with Appro to deliver this world-class system to LLNL,"said Peter Waxman, vice president sales, North America at Voltaire. "Voltaire's DDR InfiniBand solutions enable LLNL to reap the additional processing power provided by dual-core machines to be able to run complex calculations and scientific simulations faster and more efficiently." Two of the three Linux clusters will be used in the Laboratory's unclassified environment as a multi-programmatic and institutional (M&IC) resource. The other Linux cluster will be used in the classified environment to solve complex computational problems related to the Laboratory's National Nuclear Security Administration's (NNSA) program to ensure the safety, security and reliability of the nation's nuclear deterrent without testing. This program is known as the Stockpile Stewardship. Identical scalable units based on twelve, Voltaire 24-port Grid Switch(TM) ISR 9024 DDR switches with 144 quad socket nodes and 1,152 cores each will be grouped together with multiple Voltaire 288-port Grid Director(TM) ISR 9288 DDR switches to form the three Linux clusters. Multiple organizations and programs within the Laboratory will share these clusters for large, medium and small scale scientific simulations. "Peloton scalable unit deployments are the next stage in Linux cluster innovation that will enable Livermore's M&IC next generation scientific simulation capability in support of our national security missions," said Mark Seager, ICCD ADH for Advanced Technology. "With the Peloton project, the Laboratory will deploy InfiniBand interconnect into large scale production resources for the first time. Voltaire's DDR solution will significantly increase the bandwidth and messaging rate required to meet the ever-increasing demand for the large scale simulations indispensable to advancing all scientific disciplines. This DDR network is well-balanced with Quad socket, dual core Opteron nodes for our very demanding mission-driven scientific simulations and allows us to build very large Linux clusters with improved MTBF characteristics." "We are very pleased to work with Voltaire's 20 Gbps InfiniBand interconnect to power Appro XtremeServer Supercomputing clusters to meet the staggering demand for 50-100 teraFLOP/s scale scientific simulations," said Daniel Kim, CEO of Appro. "Appro delivers a totally managed HPC solution that can incorporate any of our full range of high-performance server platforms and technologies building a cost-effective, high-bandwidth clusters today on an infrastructure that is flexible and easy to scale."