About Quadrics
Customers
Presentations
Contact Details
Employment
Overview
QsNet
QsTenG
RMS
Linux Software
QsEm for Sky-X & Sky-Y
HPC Services&Solutions
Overview
Software Download
Documentation
Platform Compliance
3rd Party Tools and Applications
External Resources
Gnats
FAQ
Case Studies
Performance Results
Benchmarks
Features & Benefits
Documents Library
Overview
QsNetII E-Series Pricing System
Partners
Channel Partners
Configurations
 
Home   Screensaver   Legal   Login
 
   
   
 

MCR

In order to meet the exacting requirements of this program, Quadrics developed a custom version of its QsNet switch and expanded the number of nodes in the fat tree topology for a given number of switch stages to produce a more cost effective network. This allowed the proven supercomputer class technology of QsNet to be combined with commodity processing elements. In the collaboration, Quadrics focuses on the core enabling technologies to achieve capability within a production environment allowing LLNL to exploit the widespread availability of third-party hardware, software and service bases.
"The type of Science and Engineering calculations required by Livermore's national security mission require a cluster of this size and a very high bandwidth and low-latency interconnect. The long-term collaboration between Quadrics and LLNL has been essential in providing demonstrable and scalable performance," said Dr. Mark Seager, LLNL's Asst. Dept. Head for TeraScale Systems. "The 11.2 teraFLOP/s Dual Xeon system will significantly expand the computing resources available to Livermore's unclassified researchers."



Features

  • 1152 Dual Xeon nodes (2x 2.4GHz ) LLNL CHAOS Linux distribution
    Peak 11.6 Tflop/s
    LINPACK performance 7.634GF

  • Customized 1024-port single rail QsNet network
    (320/400 MBytes/sec Uni/Bi-directional Performance)
    4.5 us MPI PingPong
    ~2 us DMA Latency

  • Integrated Visualization and Lustre Parallel File System
    (200 MBytes/sec Client-Side Bandwidth)

  • Separate Control Network

  • Quadrics RMS and LLNL SLURM Resource Management


  • The customer's point of view

    To meet the growing needs of all of Livermore's programs for high-performance computing, the Laboratory acquired a new unclassified supercomputer, the Multiprogrammatic Capability Resource (MCR) machine, which complements the computing resources made available to the Stockpile Stewardship Program through NNSA's Advanced Simulation and Computing (ASCI) program. The MCR supercomputer was delivered during summer 2002 and became fully operational by October 31, when it was named the world's fifth fastest computer by Top 500. The machine dramatically increases Livermore's unclassified computing capability. It is being used to support important projects in biology, materials science, lasers, and atmospheric science. Classified computing is accomplished on the Laboratory's more powerful ASCI White machine.
    Using Linux cluster architecture, MCR provides 2,304 processors capable of performing 11.2 trillion calculations per second. Built by Linux NetworX and Quadrics, it is the first Linux-based supercomputer to be ranked in the top 10. At a cost of less than $14 million, MCR is a factor of 10 less expensive than the other top-ranked supercomputers. In the "operations per dollar" category, MCR ranks number one.
    http://www.llnl.gov/annual02/pdfs/national.pdf



    > Legal