GOVERNMENT

Forschungszentrum Jülich in Germany has chosen Bull to supply a 100 Teraflops-capacity supercomputer to host applications for the European Union's Fusion project.

The main aim of the EU Fusion project is to speed up research into nuclear fusion - a high-potential way of generating energy which should, in the long-term, enable us to meet the challenges both of sustainable development and the disappearance of fossil fuels.

From the moment it is deployed, the Bull HPC-FF supercomputer will be used to validate the very latest nuclear fusion computer simulation models. It will enable research in the area of plasma turbulence, one of the major challenges confronting physicists today. The new supercomputer will also be used in the areas of fast particle physics, which dominates plasmas in thermonuclear combustion, and materials physics.

"The Jülich Supercomputing Centre is proud to host and operate the Bull HPC-FF for the benefit of the fusion community. Our operation and support teams will assure the most effective usage of this best-of-breed technology delivered by Bull," said Thomas Lippert, leading scientist and director of the Jülich supercomputing centre.

Didier Lamouche, Bull Chairman and CEO said, "We are honoured by the Jülich Research Center's decision, which means Bull will be a major contributor to the European Fusion Community. This represents a major leap forward in the development of a complete European ecosystem in computer simulation."

The Bull HPC-FF supercomputer will be a key component in the preparation of the IFERC (International Fusion Energy Research Center) project: an international Data Center being established as part of a collaboration between Europe and Japan in relation to the ITER program. In particular, the Bull HPC-FF supercomputer will enable the Fusion community to prepare for using a Petaflops-scale supercomputer destined to equip the IFERC in years to come. The simulations it will carry out will enable the models developed by researchers to be refined, and will guarantee the ITER can be 'utilized' under optimum conditions: a vital imperative given that each ITER experiment will cost in the region of €500,000.

"The HPC-FF supercomputer, whose usage will be organized under the terms of the European Fusion Development Agreement (EFDA) will enable us to make progress on a number of key scientific questions, and to speed up research into nuclear fusion," Jérôme Paméla, Leader of EFDA explained. "These new resources will position Europe amongst the scientific leaders when it comes to supporting the ITER project."

An integrated computing platform delivering over 300 Teraflops

The Fusion community will not only have access to over 100 Teraflops of power provided by the new HPC-FF supercomputer, but it will also be able to tap into additional power on demand from the Bull JuRoPA supercomputer, ordered from Bull by the Jülich center in 2008. The two supercomputers- Bull HPC-FF and Bull JuRoPA - will be installed by Bull during the first quarter of 2009, to create a computing platform delivering more than 300 Teraflops overall, which will position it as one of the group of most powerful supercomputers in Europe.

"The HPC-FF computer opens the way to substantial progress in several fields of research into fusion controlled by magnetic confinement; including turbulent transport, magnetohydrodynamic instabilities, plasma/wall interaction, heating systems and materials modelling. This is important to prepare for the ITER scientific program under the best possible conditions. Indeed the modelling of ITER plasmas requires very fine grids, and this requires very extensive computer resources." explained Xavier Garbet, Research Director, and member of the HPC-FF office.

"Also the design of the DEMO demonstrator, which is the next step after ITER, will lead to the development of new materials. Modelling these materials will also require very powerful computing tools. Moreover the HPC-FF will involve experts in applied mathematics and computer sciences, whose skills are needed for the development and use of the codes capable of running of a machine of this class. Finally the HPC-FF will be useful to prepare the European scientific community for the future IFERC computer, which will be built at Rokkasho (Japan) and is due to go into operation in 2012."

Designed to support a wide variety of applications

The Bull HPC-FF supercomputer has been designed to support a wide variety of applications. It will include 540 Bull NovaScale R422 E2 servers, featuring the next generation of Intel Xeon processors. The configuration includes 1,080 processing nodes giving a total of 8,640 computing cores, each with 3GB of memory. The high density of the technologies used will enable the entire supercomputer to be housed in just 20 racks. Liquid cooling doors, developed by Bull, provide efficient management of thermal dissipation at an optimum cost compared with more traditional, air-cooling mechanisms.

The 1,080 computing nodes are interconnected via a network based on InfiniBand QDR (Quad Data Rate) technology, delivering some of the fastest connections on the market. A DirectData Networks 300TB storage sub-system is also linked to the new configuration

This new contract confirms Bull's ability to meet European challenges in high-performance computer simulation. It also affirms Bull's position as a major player in this area, particularly as a result of its partnerships with other players in the IT industry and from the world of research. Having successfully predicted and anticipated the changes in computer simulation technologies a number of years ago - particularly the growing importance of standard components and open architectures - Bull is now positioning itself as a leader in this field, both as designer and a systems architect, as a hardware supplier and systems integrator. This means the company is a remarkable force for European industry, delivering the greatest benefits for all its customers.