Europe's current position in HPC has to be enhanced. 75% of the European HPC power is within PRACE, according to Prof. Dr. Achim Bachem. This turns PRACE into the European access to HPC technology. We should not forget that a number one HPC system provides six years of technology advantage compared to an average system. After twelve years the power is on the desk.

The computational science infrastructure in Europe is still lagging behind. The European Roadmap for Research Infrastructures is the first comprehensive definition at the European level. Research Infrastructures are one of the crucial pillars of the European Research Area. A European HPC service has a foreseen impact of strategic competitiveness, attractiveness for researchers, and allows to support industrial development.

The ESFRI vision for a European HPC service looks as follows:the European HPC facilities should be at the top of an HPC provisioning pyramid. As such, Euruope needs 3 to 5 Tier-0 Centres. The Tier-1 Centres are the national centres and the Tier-2 Centres are the regional and university centres.

The estimated costs for ESFRI are pretty high, warned Prof. Dr. Bachem. Unlike the other European Research Infrstructures, the Tier-0 resources have to be renewed every 2-3 years. The construction cost amounts to 200-400 million euro every 2-3 years. The annual running cost should be rated at 100-200 milion euro. Needless to say that we face a truly European challenge, also in terms of funding, since this cannot be supported by one country alone.

The first steps and achievements within PRACE will be the production of the HPC part of the ESFRI Roadmap. The creation of this vision involves 15 European countries and will bring scientists across Europe together. The HET initiative addresses the scientific case, including domains such as weather, climatology, Earth science, astrophysics, and so on. Supercomputing drives science through simulation in environment, the ageing society, and materials science.

Within PRACE a MoU with the 15 countries involved has been signed and an FP7 project proposal has been submitted and approved. The objectives are to prepare the contracts to establish the PRACE permanent Research Infrastructure as a single Legal Entitiy in 2010 including governance, funding, procurement and user strategies. PRACE also aims to perform the technical work to prepare the operation of the Tier-0 systems in 2009/2010 including deployment and benchmarking of prototypes for Petaflops systems and the porting, optimising, and peta-scaling of the applications.

The PRACE partners include 16 legal entities with principal and general partners. The next tasks are multiple in order to grow into a persistent Research Infrastructure. In this regard, the partners have to define the legal form and governance and secure initial and continuous funding. An important issue is to prepare the procurement and installation of the first Petaflops systems and to establish the peer review process for academic use. PRACE will also promote Europe wide collaboration between scientific communities using leading edge scientific simulation and encourage new projects to increase software and simulation competence. Training and education will be provided as well.

The model for the Research Infrastructure provides a Principal Partners Committee and is built on principal and general partners, as stated by the speaker. Scientific Communities are being established headed by a Scientific Steering Committee.

In addition, ERI is a new European Legal Framework and the ESFRI roadmap 2006 foresees 35 research infrastructures. The HPC Ecosystem links to the user communities such as for example the ESFRI infrastructures, the national and European funding agencies, and the HPC hardware and software industries. There will be interoperability with Grid infrastructures. The DEISA project is a future pillar of PRACE as well as HPC-Europe.

The speaker emphasized the importance of involving the European Industry. The usage of HPC technology in the industry is 6-8 years behind the technology frontier, as available to top research. The USA undertakes to boost competitiveness of the local industry by shortening this period. In this regard, the First PRACE Industrial Seminar, to be held in Amsterdam, September 3-4, 2008 will be a major event. The automotive and aerospace industry are interested, as well as the finance and insurance sectors.

In order to face the challenge, the PRACE partners will have to identify the architectures and vendors capable of delivering Petaflops systems by 2009/2010. Another task is to install prototypes at the partner sites to verify their viability. The partners will have to define consistent operation models and evaluate management software as well as capture application requirements and create a benchmark suite. Then it is necessary to port, optimize and scale selected applications. To accomplish this, PRACE has to define an open and permanent procurement process; implement a strategy for continuous PC technology evaluation and system evolution within the Research Infrastructure; and foster the development of components for future multi-petascale systems in Europe, according to Prof. Dr. Bachem.

It is thus of utmost importance to foster the European HPC industry. Currently, most HPC vendors are US- or Japan-based. An independent access to HPC technology is a strategic issue for Euorpe. PRACE will foster European developments by translating user requirements to architectural specifications for future multi-petascale computing. The PRACE project will end in 2009, but there is also a French-German HPC co-operation in the form of a strategic partnership of the Forschungszentrum Juelich and CEA, known as Gauss and GENCI.