dcsimg
RealTime IT News

Weather Service Turns to IBM for Supercomputing

The hurricane season is about to get in full swing, and the National Weather Service says it's in better position to predict the potentially disastrous weather patterns as a result of bringing online a new supercomputer from IBM .

The new incredibly fast supercomputer begins its work today and is capable of as much as 7.3 trillion operations per second, and could attain up to 100 trillion operations per second by 2009. The IBM supercomputer for the NWS is comprised of 1,408 processors in clusters of 44 very powerful linked systems, and is located at the IBM computing center in Gaithersburg, Maryland.

NWS said the massive calculation potential will enable it to improve its predictive analysis of current and future weather patterns, and has said it will spend close to $224 million over the next nine years for the entire IBM hardware, software and service solution. The information and analysis generated by this system will be the basis of local, national and international weathercasts on television, radio and other media.

Weather experts expect the new supercomputing system will result in forecasts that will be more reliable over a five-day period, as opposed to previous technology which was usually accurate over a three-day period.

"Advanced supercomputing technology such as this system allows for weather forecasts to be more accurate, so an average citizen can better decide where to plan their weekend getaway," said Peter Ungaro, vice president of IBM Deep Computing. Ungaro said the technology effectively will allow weather forecasts to more precisely pinpoint specific weather systems.

IBM Deep Computing is a separate unit within the computing giant which specializes in high performance computing designed for weather institutions, advanced research centers, digital media companies, petrochemical, pharmaceutical and biotechnology companies worldwide. Ungaro said his group is targeting other companies that specialize in business intelligence, data mining and data warehousing.

Ungaro said "most large countries in the world have major weather centers, and we are applying similar supercomputing solutions for a number of countries around the world, including the European Center for Medium Range Weather Forecasts in Reading, England, as well as in Germany, Canada, Turkey, Hungary, Tunisia, Ireland, China, Morocco and Thailand, for a total of more than twenty international weather clients."

With these improved predictive powers, weather officials will be in a better position to track potentially deadly hurricanes and other storm systems, which can save lives, if local residents are able to find out about massive storms before they strike.

"For instance, in severe weather situations local officials have to make costly and tough decisions regarding evacuation of an island or a specific area, and these systems allow for those decisions to be made in a timely manner," Ungaro added.

During the first stage of the supercomputer rollout, which started in May, a cluster of 44 IBM p690 Regatta servers along with 42 terabytes of IBM FastT500 storage servers were deployed. This initial system has 7.3 teraflops of computing power. IBM said the final system, which will be ready in less than a decade, will likely be capable of more than 100 teraflops. A teraflop refers to a trillion calculations per second.

In addition to millions of households across the country being able to rely more on the accuracy of the local weathercast, the new technology could impact everything from military movements to the shipment of goods to and from areas of the world which often experience difficult weather systems, such as the Caribbean during hurricane season.

Ungaro said supercomputer weather modeling makes it possible for "more physics to go into the models, which in turn, allows for more accurate and faster forecasts." He said weather predictions are now using "ensemble forecasting," which allow them to run the same forecast with multiple types modifying variables to better predict the exact weather patterns.