Looking to match the kind of computing power it created in the BlueGene/L supercomputer — a system made up over more than 130,000 processors — IBM launched this week what it called the fastest privately-owned supercomputer in the world. The system is sister to BlueGene, dubbed “BGW,” and is expected to rank among the world’s three fastest public/private super systems.
Big Blue said BGW, which stands for Watson Blue Gene system, will be used to push progress in life sciences, hydrodynamics, materials sciences, quantum chemistry, molecular dynamics, fluid dynamics and even business applications.
IBM leads the charts of the Top 500 Supercomputer list with its clustered systems created from a number of lower-cost computers and processing power. The company said this approach makes supercomputing more widely available and applicable to a broader set of applications. Analysts, however, are skeptical of supercomputing’s pervasiveness, indicating only a few industries have the need for such power, and even fewer companies are willing to pay for it.
Smaller Size, More Power
IBM said BGW, developed earlier this year and already in use by IBM researchers, has achieved a data processing speed of 91.29 teraflops and is expected to join its “sister machine,” the BlueGene/L at Lawrence Livermore Laboratory. At 360 peak teraflops, BlueGene/L is the world’s reigning fastest supercomputer, but both systems are works in progress, constantly being expanded.
BGW is currently comprised of 20 refrigerator-sized racks installed at the IBM Thomas J. Watson Research Center in New York. Each rack carries a couple thousand PowerPC processors, and the whole thing takes up less than half the space of a conventional system of comparable power — yet BGW delivers three times the performance, according to IBM.
Big Blue said its advances in size, cost and power consumption make high-performance supercomputing more widely available for a broader set of applications — something prohibited by the large size and cost of yesterday’s supercomputers.
Bioscience and Business
The new supercomputer will be used by IBM to explore how such systems might enable progress in traditional scientific and technical applications of supercomputing power, including Blue Matter, a software framework developed with BlueGene that runsprotein dynamics simulations key to drug development.
The scientific research gives BGW access to large-scale research projects from universities and other research institutions. Additionally, IBM said the supercomputer would also be utilized by its Center for Business Optimization. The CBO is a consulting and software unit aiming to tap IBM’s math and industry computing expertise to tackle difficult business problems, such as high-precision weather forecasting, predictive models for disaster response, utility supply/demand forecasting, transportation, and financial markets analysis.
“IBM researchers will use BGW to accelerate discovery in a variety of disciplines,” said IBM Research vice president of systems Tilak Agerwala in a statement. “Researchers, scientists, engineers and inventors can now ask more questions, test more theories, try more designs and simulate more conditions than has been possible before.”
Limited To Niche
Gartner research vice president Martin Reynolds told TechNewsWorld that, though the smaller systems that make up today’s supercomputers are less expensive, the ultra high-end computing still comes at a price and is not necessary for most enterprise players.
“It’s a small market,” Reynolds said. “These [systems] generally go in places where you need massive amounts of calculations, and they’re expensive because there’s a lot of silicon in them.”
Reynolds said some businesses are taking advantage of the ability to link up smaller systems for the power of a single, more expensive solution. Nevertheless, he emphasized that, beyond weather and financial market simulations, there are few businesses willing to pay for this kind of computing power.
“In terms of enterprise, it’s a matter of whether there is enough data [to demand a supercomputer],” he said. “To some extent, that’s offset by good managers.”