SCIENCE
Appro Deploys Supercomputing Cluster to LLNL in Support of Interactive Data Analysis and Visualization Project
Appro today announced the deployment of Appro Hyper Series Supercomputing Clusters to provide Lawrence Livermore National Laboratory (LLNL) with a new visualization cluster; called “Graph” geared specifically to support interactive data analysis and visualization on the extreme scale Sequoia classified computing systems.
The lab visualization resources provide an essential and critical part of the high performance computing environment enabling computational scientists to interact and explore very large data sets as well as large-scale post-processing and data reduction operations. Development of these capabilities is vital to the National Nuclear Security Administration’s (NNSA) program to ensure the safety, security and reliability of the nation’s nuclear deterrent without underground testing – stockpile stewardship. This HPC project required a memory-rich computing with I/O optimized through dedicated file system connections.
Appro deployed to LLNL four Scalable Units (SU)s – 110 TeraFLOP/s Linux clusters. The Graph cluster consists of 564 compute nodes, 2,256 processors / 13,536 cores with 73 TB of memory. Each compute node is based on Six-Core AMD Opteron Processors (code named “Istanbul”) at 2.0GHz with 128GB of system memory, 48 Flextronics InfiniBand 20 Gb/s DDR edge switches, and 2 Voltaire DDR Spine switches.
“To meet our current and future scientific computing needs requires a visualization cluster with enough memory to generate visualizations of simulation runs on our largest compute platforms as well as sufficient I/O rates for interactive analysis,” said Becky Springmeyer, Visualization group leader in LLNL’s Advanced Simulation and Computing Program.
The largest machine to be supported by Graph initially will be Dawn, a 500 teraFLOP/s system delivered earlier this year, followed by Sequoia, a multi-petaFLOP/s system to be delivered in 2011. Sequoia will have 1.6 petabytes of memory, 98,304 compute nodes and 1.6 million cores. These supercomputers are capable of running very large suites of complex simulations and Graph will be capable of supporting complex visualization and analysis tasks on data sets generated on these much larger machines. Visualization specialists are dealing with multi-terabyte data sets with tens of billions of zones, thousands of files per time step, and hundreds of time steps. Post-processing tasks are heavily I/O bound, so specialized visualization servers that optimize I/O rather than CPU speed are better suited for this work. These simulations will be now enabled through Graph.
"The higher end of the HPC market is one of the faster-growing HPC segments and IDC projects that it will continue to see healthy growth over the next 3 to 5 years as more countries and organizations apply very large HPC systems to address leading scientific and engineering problems," said Steve Conway, IDC Research Vice President for Technical Computing. "Appro HPC clusters are designed to give LLNL users significantly more memory and high-speed connections to Lustre than they have on previous LLNL clusters of comparable size. This means that visualization problems can finish faster and more users will be able to post and process their data simultaneously.”
“Appro is proud to be able to provide Lawrence Livermore National Laboratory with a powerful visualization cluster that will enable important national security work and push the state-of-the-art in scientific computing on some of the world’s most powerful high performance computing systems,” said John Lee, VP of Advanced Technology Solutions of Appro.
“Six-Core AMD Opteron processors enable superior performance scalability, helping to allow HPC environments, like the Graph Cluster, to achieve outstanding results,” said John Fruehe, director, Business Development, Server and Workstation Division, AMD. “AMD’s Direct Connect Architecture and HyperTransport technology with HT Assist help deliver the node memory bandwidth and support for large-scale memory footprints needed for efficient data analysis and complex visualization tasks.”
Appro is a leading developer of supercomputing solutions. Appro is uniquely positioned to support High Performance Computing markets focusing on small to large-scale deployments where lowest total cost of ownership is a primary consideration. Appro accelerates technical applications and business results unlocking the value of IT through outstanding price/performance, balanced architecture, open standards and engineering expertise. Appro headquarters is in Milpitas, CA with offices in Korea and Houston, TX. To learn more go to http://www.appro.com
The lab visualization resources provide an essential and critical part of the high performance computing environment enabling computational scientists to interact and explore very large data sets as well as large-scale post-processing and data reduction operations. Development of these capabilities is vital to the National Nuclear Security Administration’s (NNSA) program to ensure the safety, security and reliability of the nation’s nuclear deterrent without underground testing – stockpile stewardship. This HPC project required a memory-rich computing with I/O optimized through dedicated file system connections.
Appro deployed to LLNL four Scalable Units (SU)s – 110 TeraFLOP/s Linux clusters. The Graph cluster consists of 564 compute nodes, 2,256 processors / 13,536 cores with 73 TB of memory. Each compute node is based on Six-Core AMD Opteron Processors (code named “Istanbul”) at 2.0GHz with 128GB of system memory, 48 Flextronics InfiniBand 20 Gb/s DDR edge switches, and 2 Voltaire DDR Spine switches.
“To meet our current and future scientific computing needs requires a visualization cluster with enough memory to generate visualizations of simulation runs on our largest compute platforms as well as sufficient I/O rates for interactive analysis,” said Becky Springmeyer, Visualization group leader in LLNL’s Advanced Simulation and Computing Program.
The largest machine to be supported by Graph initially will be Dawn, a 500 teraFLOP/s system delivered earlier this year, followed by Sequoia, a multi-petaFLOP/s system to be delivered in 2011. Sequoia will have 1.6 petabytes of memory, 98,304 compute nodes and 1.6 million cores. These supercomputers are capable of running very large suites of complex simulations and Graph will be capable of supporting complex visualization and analysis tasks on data sets generated on these much larger machines. Visualization specialists are dealing with multi-terabyte data sets with tens of billions of zones, thousands of files per time step, and hundreds of time steps. Post-processing tasks are heavily I/O bound, so specialized visualization servers that optimize I/O rather than CPU speed are better suited for this work. These simulations will be now enabled through Graph.
"The higher end of the HPC market is one of the faster-growing HPC segments and IDC projects that it will continue to see healthy growth over the next 3 to 5 years as more countries and organizations apply very large HPC systems to address leading scientific and engineering problems," said Steve Conway, IDC Research Vice President for Technical Computing. "Appro HPC clusters are designed to give LLNL users significantly more memory and high-speed connections to Lustre than they have on previous LLNL clusters of comparable size. This means that visualization problems can finish faster and more users will be able to post and process their data simultaneously.”
“Appro is proud to be able to provide Lawrence Livermore National Laboratory with a powerful visualization cluster that will enable important national security work and push the state-of-the-art in scientific computing on some of the world’s most powerful high performance computing systems,” said John Lee, VP of Advanced Technology Solutions of Appro.
“Six-Core AMD Opteron processors enable superior performance scalability, helping to allow HPC environments, like the Graph Cluster, to achieve outstanding results,” said John Fruehe, director, Business Development, Server and Workstation Division, AMD. “AMD’s Direct Connect Architecture and HyperTransport technology with HT Assist help deliver the node memory bandwidth and support for large-scale memory footprints needed for efficient data analysis and complex visualization tasks.”
Appro is a leading developer of supercomputing solutions. Appro is uniquely positioned to support High Performance Computing markets focusing on small to large-scale deployments where lowest total cost of ownership is a primary consideration. Appro accelerates technical applications and business results unlocking the value of IT through outstanding price/performance, balanced architecture, open standards and engineering expertise. Appro headquarters is in Milpitas, CA with offices in Korea and Houston, TX. To learn more go to http://www.appro.com