Server Specs: The week in supercomputing

IBM dominates the Top500 project; Cisco unveils new products for InfiniBand; Sun looks to go big in Japan; and Linux Networx attempts HPC for the masses.

Supercomputing giants gathered this week in Seattle for their annual conference, Supercomputing 2005, to discuss trends and unveil new products. Here is a compilation of news from the event.

Big Blue dominates Top500

This week, the Top500 project announced its latest biannual list of the most powerful supercomputers in the world, and IBM took the top three consecutive spots and five of the top 10 overall.

In this 26th iteration of the list, IBM's BlueGene/L at the Lawrence Livermore National Laboratory in Livermore, Calif., retained the top spot, hitting 280.6 teraflops, or trillions of floating point calculations per second on the Linpack performance scale -- a performance metric from the Top500. No other supercomputer has ever achieved more than 100 teraflops.

Lawrence Livermore National Laboratory is part of the National Nuclear Security Administration within the Department of Energy, and has been managed since its inception in 1952 by the University of California. The laboratory is responsible for ensuring that the nation's nuclear weapons remain safe.

The Top500 is compiled and published by supercomputing experts Jack Dongarra from the University of Tennessee, Erich Strohmaier and Horst Simon of NERSC/Lawrence Berkeley National Laboratory and Hans Meuer of the University of Mannheim (Germany). The entire list can be viewed here.

For more information:

Supercomputers come off the shelf, skills don't

Learning Guide: Grid computing

Cisco unveils new products around InfiniBand Cisco Systems Inc. unveiled some major products for the supercomputing market this week at the Supercomputing 2005 conference. The networking giant broke into the supercomputing business with its acquisition of Topspin Communications, Inc. earlier this year. The first Cisco products from the acquisition became available in September. The new product releases include:

  • Cisco SFS Subnet Management Software, a network management tool designed to discover network nodes and route traffic across large scale supercomputing clusters.
  • 144-port and 288-port Cisco server fabric switches.
  • Cisco Superflex InfiniBand flexible copper cables, which are more bendable than traditional InfiniBand connections and can run InfiniBand over a wide area network of up to 50 miles.

    The announcements coincide with Cisco's first showing in the top 10 of supercomputing's ever-changing who's who list, the Top500. Cisco customer, Livermore, Calif.-based Sandia National Laboratory, took fifth and sixth on the list with two powerful systems.

    Sun looks to be big in Japan Sun Microsystems Inc. announced plans to build what it claims will be Asia's most powerful supercomputer. Sun has partnered with Advanced Micro Devices and NEC Corp. to build a Sun Fire-based cluster for the Tokyo Institute of Technology. This new system is scheduled to be operational by 2006.

    Sun was absent from the top 10 positions on the Top500 for this announcement, but claims that this system, when completed, will hit over 100 teraflops, a number reached only by the current supercomputing leader, the BlueGene/L at the Lawrence Livermore National Laboratory.

    Satoshi Matsuoka, a professor in charge of research infrastructure at the Tokyo Institute of Technology, said researchers plan to use the system for tackling several problems, including analyzing the molecular structure of proteins, simulated blood flow diagnosis in human brains and simulation of carbon nanotubes -- all tasks that require exceptional computing power.

    Linux Networx attempts HPC for the masses

    Bluffdale, Utah-based supercomputing company Linux Networx unveiled two new lines for the high performance computing (HPC) market, one of which it claims will break down the barriers for commercial companies interested in supercomputers.

    Linux Networx said its LS-1 was specifically designed to make supercomputing more affordable and manageable for midrange users. The company is providing application-specific configurations for companies facing science and engineering challenges.

    According to the company, no special facilities or cluster expertise is required, and the systems are ideally suited for off-the-shelf, third-party applications for computational fluid dynamics, crash analysis, structural engineering, flow simulation and interactive 3D visualization.

    The other line being introduced, the LS/X, is an ultra high-end, fault-tolerant Linux supercomputer designed to avoid bottlenecks common in first generation clusters.

    Let us know what you think about the story; e-mail: Matt Stansberry, News Editor

  • Dig Deeper on Colocation, hosting and outsourcing management

    Start the conversation

    Send me notifications when other members comment.

    Please create a username to comment.