Weather service leases IBM supercomputer
The National Oceanic and Atmospheric Administration's National Weather Service (NWS) is leasing a new supercomputer from IBM Corp. that is so powerful that the service will eventually be able to forecast the weather neighborhood by neighborhood.
During the first three years of operation, the new supercomputer will provide almost five times the computational power of NWS' current IBM supercomputer. With upgrades over the duration of the service's nine-year, $224.4 million contract, the supercomputer will reach 48 times the computational power of the current computer by October 2009, NOAA officials said.
"Currently, we can do a grid point forecast for New York City," said Carl Staton, NOAA's chief information officer. "By the end of the contract period, it's possible that we'll be able to do a grid point forecast for Queens and a grid point forecast for Manhattan."
NOAA decided to lease, rather than purchase, its new supercomputer because the facility in Bowie, Md., that houses the current supercomputer can't accommodate the new machine without major upgrades to its floor space and cooling systems, Staton said.
"The Census Bureau is the primary occupant of that building now, and they have their own requisites," he said. "The co-location of our computer at that facility with Census doesn't work for us."
Instead, the new supercomputer will be housed at IBM's Gaithersburg, Md., facility. Installation of the new machine will be completed by Sept. 30, and the transition of operations to the new system will be completed by the end of July 2003.
Even though the new supercomputer won't be housed in a government building, NOAA researchers will have high-speed access to it "so, for them, it will be like it is in their own facility," said Peter Ungaro, vice president of high-performance computing in the public-sector division of IBM.
The new supercomputer will actually be a series of about five machines installed over the course of the contract period, as IBM adds new technology and computing power, he said.
Staton doesn't anticipate problems in using the new supercomputer remotely. "We see no immediate downsides in having it at the IBM facility. We're very excited about this new approach," he said.
The new supercomputer's increased processing capabilities will enable NWS to significantly improve weather, flood, ocean and climate forecasts for the country, said retired Navy Vice Adm. Conrad Lautenbacher Jr., undersecretary of the Commerce Department for oceans and atmosphere and NOAA administrator, when the lease signing was announced late last month.
"The accuracy of environmental forecast models today is approaching levels undreamed of 10 years ago," Lautenbacher said. "As a result of this new supercomputer, the National Weather Service can run more powerful models with improved physics to produce forecasts with better resolution, accuracy and lead times than ever before."
The new supercomputer, which uses a highly parallel computer architecture with 2,752 processors, will include a data archive that can pull more information from NWS satellites than ever before, Ungaro said.
It also will be powerful enough to incorporate vast amounts of data even if they are changing literally moment to moment as threatening weather conditions develop, said Louis Uccellini, director of NWS' National Centers for Environmental Prediction in Camp Springs, Md.
"The production of numerical predictions within the operational time windows required by field forecasters is one of the most computationally intensive undertakings ever attempted within an operational environment," Uccellini said.
The speed and computing capabilities of the new supercomputer will allow NWS to predict weather conditions from five to seven days — and even 10 to 14 days in some cases — in advance, NOAA officials said.
Once the new supercomputer is installed, it will be one of the largest machines of its kind in the world, only slightly behind the ASCI White supercomputer at Lawrence Livermore National Laboratory, Ungaro said.
NEXT STORY: DOT seeks smart card tech