Entries filed under “GPUs”

Our series of features on European HPC vendors continues with this interview with Pascale Bernier-Bruna from Bull. The company recently deployed the hybrid 2 Petaflop CURIE supercomputer, the first French Tier0 system open to scientists through the French participation in the PRACE research infrastructure.

insideHPC: What is the current status of CURIE, the 2 Petaflop supercomputer designed by Bull for GENCI? Is it fully deployed and running customer workloads?

Pascale Bernier-Bruna: The CURIE supercomputer – which was implemented in two stages between late 2010 and the end of 2011 – is now completely installed and is fully available to the scientific community since the beginning of March 2012.

In its final testing phase from December 2011 to February 2012, CURIE’s effective operation has been verified by running a number of very large-scale simulations using virtually all of its components. This so-called “Grand Challenges” phase also enabled researchers to achieve major scientific advances.

For example, this was the case with the work carried out in December 2011 by the team led by Michel Caffarel from the quantum chemistry and physics laboratory at CNRS/Paul Sabatier University in Toulouse. In order to gain a better understanding of the chemical phenomena at work in the process of neuro-degeneration – particularly in Alzheimer’s Disease, which currently affects over 20 million people worldwide, researchers were looking to model the behaviour of metallic ions that are particularly involved in these processes.

The simulations carried out using virtually all CURIE’s processing cores with the QMC=Chem code proved to be highly more accurate that those obtained so far using classical methods.

In astrophysics, a team of researchers from the Paris Observatory, coordinated by Jean-Michel Alimi, has performed the first-ever computer model simulation of the structuring of the entire observable universe, from the Big Bang to the present day. The simulation has made it possible to follow the evolution of 550 billion particles. This is the first of three runs which are part of an exceptional project called “Deus: full universe run”, carried out using CURIE.

This simulation, along with the two additional runs expected by late May 2012, will provide outstanding support for future projects dedicated to the observation and mapping of the universe. They will shed light on the nature of dark energy and its effects on cosmic structure formation, and hence on the distribution of dark matter and galaxies in the universe.

The implementation of “Deus: full universe run” represents a new stage in the development of supercomputing. The first simulation in the project has largely outperformed the most advanced cosmological simulations carried out over the past few years by a number of international collaborations at the largest supercomputing facilities around the world. The entire project will use more than 30 million hours (about 3500 years) of computing time on virtually all CPUs of CURIE.

Other research teams have high hopes of CURIE, including those from the CEA working on nuclear fusion, with the aim of scoping the future prototype of ITER (the International Thermonuclear Experimental Reactor). Researchers at CORIA and CERFACS are planning to use the system to optimize the combustion processes in turbines and piston engines. And teams from the Pierre Simon Laplace Institute (IPSL) will be creating multi-level climate models, to study cyclones in the Indian Ocean.

insideHPC: What would you say Bull’s unique strengths are as a key vendor in the worldwide HPC market?

Pascale Bernier-Bruna: In its HPC strategy, Bull has three major components. The first one is that Bull has proved with the Tera-100, Curie and Helios systems that it can successfully design and implement petaflops-scale supercomputers integrating thousands of servers, tens of thousands of processor cores and complex storage architectures. Now when you have installed some of the most powerful supercomputers in the world, you are obviously more than capable of handling all types of projects.

And the second key point is that Bull masters every aspect of global HPC solutions design: compute node design, interconnect optimisation, appropriate software stack development, application performance monitoring and optimisation.

The third and main component is the quality of Bull’s HPC technical team. The development, engineering and support are mainly based in Paris and Grenoble (France). They are ideally located to support our European customers, in the same time zone, which improves our capacity to develop closer relationships with our customers.

Merchants of doom predicted that investing all our R&D resources on the European soil was an obstacle to Bull’s expansion in HPC, but it has in fact revealed to be an advantage in the current context.

Quite rightly, the European Commission recently reaffirmed the strategic importance of HPC to the continent, for the competitiveness of its businesses and the creation of employment. So Europe is planning to double its investment in HPC between now and 2020. And Bull – as Europe’s only manufacturer of supercomputers and a leading player with a global reputation – is determined to play a pre-eminent role in this coming drive.

insideHPC: What will you be showcasing this year at ISC’12?

Pascale Bernier-Bruna: Bull will be exhibiting its complete range of Extreme Computing solutions, based on the bullx family of systems designed specifically for HPC. The latest evolution of the bullx blade system, the B510 blades, will be in the limelight, since they are at the heart of the two petascale systems installed by Bull recently. The bullx B510 blades are suited for configurations of all sizes, not just for large-scale supercomputers, and will successfully meet the performance requirements of a large variety of HPC users, as did the previous generation of bullx blades.

Bull will also be exhibiting its coming addition to the bullx family, the bullx B700 Direct Liquid Cooling blades, which deliver drastic savings in energy consumption. Their revolutionary technology brings cooling at the heart of the blades themselves, and makes it possible to use warm water for cooling, while the systems remain extremely easy to maintain.

insideHPC: How important is the ISC conference to your HPC marketing plans on an annual basis?

Pascale Bernier-Bruna: Bull is the only European-based HPC manufacturer, ISC is the largest European-based HPC event… It is a natural fit for Bull to be a Platinum sponsor of ISC!