HPC is not an instrument or playstation for scientists but a key technology. In the USA they have realized that HPC is a key technology for international competitiveness in science and engineering. HPC is a continued endeavour. With this firm statement, Dr. Bachem welcomed the industry people.

So what is done in Europe so far? If we consider the recent TOP500 list, there was three times more capacity in the USA than in Europe in 2006. This is now only twice and 75% of the European HPC infrastructure is now represented in PRACE. This gives Europeans a chance to compete on the same level as the USA, according to the speaker.

Europe needs an independent access to this key technology just like it has to space with ARIANE. Supercomputing competence is needed not only for science but also for industry. There are examples in energy, science, automotive, aerospace, etc. and we need to connect different levels.

Dr. Bachem also pointed out that there is a 6 years technology advantage with a no. 1 system. If we look at the average power of supercomputers on the TOP500 from 1993 to 2007, we have reached 10 Teraflops in 2007. So, it is really important for industry to have the key technology ready six years in advance. Industry problems can better be solved on HPC machines. In the USA, the INCITE programme constitutes a very good example, the speaker stressed.

The computational science infrastructure in Europe has been laid down in the Roadmap for Reserach Infrastructures in Europe, known as ESFRI. This is a list of European infrastructure which is important for science. Supercomputing technology is on this list. The ESFRI Vision for a European HPC service is put into the picture of a pyramid. Not only supercomputers but also the connection to those machines in industry and academic institutions which we call national supercomputing centres, is of capital importance.

The supercomputer however is not the first environment to talk about, as Dr. Bachem pointed out. From the bottom to the top, there are tier-2, tier-1 and tier-0 centres as a picture of the ecosystem. Not only the machines are important but also the knowledge.

Why do we need a European approach? To stay on the top level worldwide, that's why. Unique resources cost 100 to 200 million euro per year. Needless to say that the country member states cannot do this alone. This has become a true European challenge.

The first steps and achievements have been a challenging task: coming together takes time. It all started out in 2004 when scientists came together to present a scientific case. In a small period of 1 year the stakeholders came together in Finland. Then there came support from the European Commission.

The scientific case is known as HET. Weather/climatology/astrophysics/plasma physics etc. all need petaflop power. Without supercomputers there are no scientific solutions, warned Dr. Bachem. The supercomputer drives science through simulation. We all know that medicine, biology and energy constitute a big challenge.

Therefore there was an urgent need to prepare the infrastructure as a single legal entity with a good deal of funding for the European customers in industry and academia. Now what exactly are the needs in industry? This is what PRACE wants to find out. The project consists of 16 legal entities from 14 countries. A Memorandum of Understanding has been signed for software issues: 20 European countries have signed the document and the project started beginning 2008 and ends next year.

Dr. Bachem pointed out that level 2 is the software: PRACE is searching for an improved performance through software technology, so not only the machines or hardware are important. More funding will be necessary for the software.

With regard to exaflop computing, Dr. Bachem explained that the systems have to scale: otherwise there will be no exaflop performance.

If we look at the current state of HPC services for the European industry, we find that there is a lack between what could be done and what is achieved today. In fact, Europe is lagging behind some 8 years, according to the speaker. The USA undertakes to boost competitiveness of local industry by shortening this period with the INCITE programme. PRACE is striving towards a similar model. What will be the impact of PRACE foreseen for European industry? Will it be sector by sector or will it involve the whole industry all together? PRACE wants to get to know the industry needs.

Dr. Bachem stated that an open discussion is necessary. Many questions are involved concerning security and privacy, for example, the usage model, and the long term commitments, not to forget.

Next tasks involve the identification of the architectures. There will be two systems in 2009, which will be installed at the partner sites. There is a need for consistent operation models. The project also has to define an open, permanent procurement process; define and implement a strategy for continuous HPC technology evaluation; and develop components for multi-petascale systems.

Dr. Bachem is very much interested in fostering a European HPC industry with European knowledge and competence. An independent competence in Europe is high on his PRACE priority list. The project partners therefore will invite all vendors to base their research in Europe and share their knowledge in European research labs. Examples up till now are TALOS, Bull, CEA, HLRS, Intel, and Quadrics. Currently there are two consortia: TALOS and Prospect.

PRACE roadmap included the DEISA, eDEISA and DEISA2 projects from 2006 to 2011. One European HPC service is the ultimate goal. Dr. Bachem definitely sees opportunities ahead. PRACE builds upon the HPC expertise of 14 European countries in HPC service provisioning and on projects like DEISA, the expressed support of the national governments, the European Commission and many scientific communities.