Since 1986 - Covering the Fastest Computers in the World and the People Who Run Them

Language Flags
July 19, 2007

European Meterological Institute Launches NEC Supercomputer

by Markus Henkel

The Central Institute for Meteorology and Geodynamics (ZAMG) in Vienna has officially launched a new vector supercomputer. Capable of performing 512 billion arithmetic operations per second, the new NEC supercomputer is 28 times faster than its predecessor, an SGI supercomputer, and is designed to generate weather forecasts more quickly and with greater precision.

On June 19th, 2007, the NEC vector computer SX-8R, which has been in use since February, was ceremonially inaugurated. The installed NEC-SX-8R consists of two nodes, each with eight processors and connected by a high speed interconnect that was specially developed by NEC for that purpose. Each processor is running at 2.0 GHz. Each node achieve 256 gigaflops, for a total of 0.51 teraflops. Dr. Georg Kaindl, manager of the ADV, which is the institute’s department responsible for the operating of the computing centre and the computer network, explained that the uniqueness of the installation was due to the institute’s demanding system requirements. Only about 12 square metres of space was available, Kaindl added, and an expected enhancement of the computing power by a factor of 25 to 30 combined with the need for low energy consumption complicated the process even more.

This was also the main reason why the vector computer was eventually selected. Kaindl admitted that the decision for the new system was eased by the sum of its advantages over its predecessor. The most striking advantages of the NEC machine are its significantly smaller size, it faster performance, and its lower power consumption. Indeed, the new model has enormously enhanced computing power and has reduced the expense of energy and cooling because of its lower heat emission.

For ten years, ZAMG has been part of the ALADIN consortium in which various countries and their weather services are striving to improve and further develop short-term weather forecasting capabilities. ZAMG is a part of the RC-LACE (Regional Centre for Limited Area Modelling in Central Europe) group within the ALADIN consortium. Croatia, the Czech Republic, Hungary, Slovakia, Slovenia and Romania have teamed up to create a scientific network in which about 100 scientists are continuously working on ALADIN-NWP (Numerical Weather Prognosis Project). The director of ZAMG, Dr. Fritz Neuwirth, claimed that weather as a field of research has become so complex that any single regional weather service finds it difficult to provide state-of-the-art service. Consequently, the established cooperation was enormously important. The ALADIN model has been especially useful in the Alps and Central Europe. The ALADIN model, which performs weather forecasting tasks along with ozone and crisis models, runs 28 times faster on the new computer than on the old machine.

The basic idea of the ALADIN project, establishing a cooperation of the national weather services of Central and Eastern Europe for a common benefit, was initiated by Meteo France in 1990. Its tasks are:

  • The operation and maintenance of an NWP system for geographically limited areas. This method only needs moderate computing capacities, but unlike the global ARPEGE model, it provides a magnifying effect.

  • Operating with small model areas and big meshes; it is presumed that the decisive meteorological elements on a fine scale, such as local winds, breezes or thunderstorm lines, are mainly caused by the dynamic adaptation of the air streams to the earth’s surface.
  • The construction of an international NWP system with high standards in which all partners participate effectively and eventually share the common result.

ZAMG’s former system only permitted a resolution of about 9.6 kilometres, which no longer allowed them to be competitive, Neuwirth stated. The new supercomputer enables ZAMG to perform a resolution of 5 kilometres, and is expected to operate with a resolution of 2.5 kilometres in the future. Combined with a reduction of the time intervals between the actual calculations, Neuwirth stated that ZAMG is now able to generate four forecasts per day, instead of the usual two, which is an important improvement in quality for the institute. The institute’s director concluded that by calculating with narrower grid points, using the aforementioned mesh amplitude of 2.5 kilometres, the international accuracy standards would be matched and both present and future cooperation could be implemented.

In the same time period as ZAMG’s upgrade, the French weather service Meteo France acquired an almost identical computer, with respect to the architecture, so that in the future the software developed in a corporate prediction model project can be used more quickly and thus more efficiently at ZAMG. Neuwirth is confident that the new machine will ease cooperation by, for example, integrating the faster AROME weather model used by the French.

With the qualitative improvement in weather forecasting, ZAMG will achieve a higher level of accuracy. This will support the ability of authorities to fulfill their legal obligations in initiating the necessary measures to prevent damage to persons or property in extreme weather situations.

Makoto Tsukakoshi, manager of NEC High Performance Computing Europe Ltd., proudly stated that the ZAMG project demonstrated NEC’s ability to install high-performance computer systems with the lowest power consumption in comparison to other systems.

Initially, ZAMG and NEC have agreed to work together for five years. They would like to extend this cooperation in the future, enabling additional progress in weather and climate research, and providing NEC developers with some direction for improving hardware and software related to the needs of the ZAMG community.

—–

About the Author

Markus Henkel is a geodesist, science writer and lives in Hamburg, Germany. He writes about supercomputing, environmental protection and clinical medicine. For more information, email him at info@laengsynt.de or visit his websites: http://laengsynt.de and http://deepwave.org.