Twitter Facebook Google Plus Linkedin email Print Pdf

|||

Bull inaugurates its Center of Excellence for Parallel Computing: the first European center of expertise dedicated to research into enhancing High-Performance Computing (HPC).

Posted on 15 March 2013 by Pierre Picard

In order to take full advantage of the power of tomorrow’s supercomputers, most HPC applications in use today will have to be modernized. This is a massive issue: it will involve constant moving forward, towards ever higher levels of innovation and competitiveness.

On March 22, Bull will inaugurate its Center of Excellence for Parallel Computing, the leading European center of technical and industrial excellence in this field. The center’s mission is to support researchers and engineers in research centers and industry to overcome the critical technological barrier of ‘HPC application parallelization’[1].

Parallel computing applications: an imperative for industry and science

In many industrial and scientific areas, applications developed many years ago are no longer suited to supporting the so-called ultra-parallel architectures (which feature thousands of processors) underpinning today’s High-Performance Computing systems. According to industry analysts, only 1% of applications in use today could run 10,000 processors or more at any one time. However, the next-generation of ‘Exascale’ supercomputers[2] will include over a hundred million processors. As a result, designers are now faced with the challenge of adapting their software if they want to benefit from the power of ultra-parallel infrastructures such as, for example, the new Intel® Xeon® processors and Intel® Xeon Phi™ co-processors. Once these new supercomputers become available, only those research centers and industries that have parallelized their applications will be able to fully reap their benefits.  The race has started… on a global scale.

Application parallelization – with smaller carbon footprint – is crucial for the development of supercomputers in the coming decades

Drawing on the expertise of a team that is unique in Europe, the Center will be equipped with the highest levels of expertise, to help research labs and companies optimize their applications so they can be compatible with the processors currently being developed. The Center will supply a broad portfolio of services, including analysis and consultancy, as well as software parallelization and optimization.

Bull’s Center of Excellence for Parallel Computing will benefit from close cooperation with Intel in the area of supercomputers and parallel computing. It will combine the very latest Intel technologies with network and storage technologies. Those using the center will also be able to benefit from partnerships with world leaders and players in research and industry specializing in parallel computing software and methodologies. Partners involved the Center of Excellence for Parallel Computing already include Allinea, CAPS and the Joseph Fourier University.

An example with the Pulsation project monitors the development of cyclones in the Indian Ocean, or How Parallel Computing is becoming a key environmental tool  

Certain regional natural phenomena across the planet can escape detection by scientific observers if the relevant computing models are inadequate.

Some of cold ocean currents create cloud cover that is impossible to identify if the supercomputer’s computational grid is too large, with the result that simulations can deliver false results.

Parallel computing can speed up the grid calculation by a factor or ten, because it delivers ten times the power, enabling very precise simulations to be built which allow observers to identify new natural phenomena (cyclones) that could never have been detected by earlier simulations. El Niño (the cold ocean current) and La Niña (the warm ocean current) are recurring regional phenomena that were particularly difficult to predict and analyze before the High-Performance Computing applications used by weather forecasters were upgraded for parallel computing.

 

Plus d’information >>> http://www.bull.com/extreme-computing/index.html


[1] Parallel computing is all about applications’ capacity to carry calculations simultaneously (hence, in parallel), without having to wait for the end of one calculation before starting another.

[2] Supercomputers with a processing power in excess of one Exaflop, or a billion billion operations per second.

Comments are closed.