Based in Budapest, HMS oversees a broad range of activities in the field of numerical weather prediction -- from ultra short-range modeling (up to a few hours) to climate-range modeling (up to several decades). In recent months, HMS administrators concluded that their existing computing resources could not keep up with the escalating demands of ultra short-range and short-range weather prediction. These forecasts have grown increasingly important throughout the world as severe weather phenomena and natural disasters have impacted lives and property.

To do this, HMS scientists gather all the observations (hourly surface measurements, vertical soundings of the atmosphere, radar data, satellite images and lightning detection data) available at the Carpathian Basin. Then in the process of numerical weather prediction, this information is combined with numerical models to form an efficient, reliable and modern forecasting system.

Using this complex forecasting system, HMS can, for instance, warn residents and businesses all over Hungary (including the popular Lake Balaton summer resort region) of potentially dangerous weather up to six hours before its arrival. The system can warn area residents of rapidly forming snowstorms, freezing rain, fog, convective storms, wind gusts, hail storms and flash floods -- and even issue automated alerts.

"Analyzing multi-source data for ultra-short range and short range prediction, and doing it quickly, has proven extremely compute- and memory-intensive," said Andras Horanyi, meteorologist at HMS.

"Our plans involve moving from our current advanced models to even higher-resolution models, which prompted the investment in a powerful and scalable parallel supercomputer to drive our codes. After extensive benchmarks involving many competing systems, SGI Altix proved the superior choice. The new Altix system offers the capabilities we require for our urgent current needs along with the headroom to scale as our needs grow over time."

SGI Altix Storms Competitors

HMS embarked on an exhaustive evaluation process prior to making its selection. The key criterion was price/performance, with performance tests focused on two codes vital to operations at HMS: ALADIN (a numerical weather prediction model developed by a broad consortium of European and North-African countries) and MM5 (the Penn State/NCAR Mesoscale Model).

Administrators at HMS evaluated systems from IBM, HP and SGI. After the benchmark evaluations, HMS selected an SGI Altix 3700 Bx2 supercomputer powered by 144 Intel Itanium 2 processors and 288GB of memory, and running Novell SUSE Linux Enterprise Server 9.

The SGI shared-memory platform so excelled in the benchmark tests that the 144-processor Altix configuration outperformed competing systems powered by approximately 200 processors and even more memory. HMS plans to install the system in two phases: 72 processors and 144GB of memory in February, and the rest of the configuration by May 2006.

"With this new Altix system, we can compute forecasts using our recent settings in just a few minutes, instead of the hour that we currently need on our existing hardware," said Horanyi. "This represents such a dramatic increase in our capabilities that it will help all of Hungary stay prepared for potentially damaging weather systems."

Related Links
Silicon Graphics