With the ETP4HPC in Europe and the HPC and Simulation call for proposals in France, European and French authorities are finally realizing the strategic role of HPC in innovation and competitiveness. In your opinion, what has led to this increased awareness?

It comes simply from the comparison with the United States, which invests millions of dollars in this sector each year. But it also results from the vigorous mobilization of the scientific community in France and Europe to alert the public authorities that our competitiveness, our innovative energy, and our international expansion are all falling behind. It took this huge communication effort to raise our leaders’ awareness of the strategic role of digital simulation.

We should also emphasize that from now on, high performance computing will be indispensable in many sectors. This is the case in health care, for example, where legislation prohibiting animal testing is a real catalyst for the simulation of living creatures and the creation of new molecules.

In this regard, how well do you think the national and European programs complement one another? In what ways aren’t they simply competing with one another?

In Europe, the commonly used term is “subsidiarity.” What can’t be achieved at the national level is achieved at the European level. For the record, the greatest allocation of hours in the world was made by PRACE on the German machine Hermit for an English team from the Met Office: no less than 144 million hours were granted to enable this team to gain three years’ lead in the development of high resolution climate models. Clearly, even our American friends had not made such an allocation. Without PRACE, this project would not have been possible. Computing cycles are allocated to projects that are on a larger scale than those that we can undertake at the national level. PRACE has a Scientific Council, an independent body that judges projects on the basis of their scientific excellence, and that made the bet to allocate its computing hours to important projects simply because the request was exceptional. This is how Europe draws its partner countries upwards and onwards, by making scientific investments that are more easily achievable at the European level than at the national level.

Let’s take a minute to focus on our governing bodies. What are the ambitions of high performance computing policies? And do you think the allocated resources are adequate with respect to these ambitions?

At the European level, Europe does have a genuine political will. This is even the first time that such a will has been displayed in the HPC area. The Commission wants to have a Europe that is competitive on the international stage, and we would hope that the Horizon2020 program will yield the results hoped for, even if, considering the €15 million allocated in the 2014-2015 Work Programme, the financial resources are not realistically adequate today. However, all is not settled, and it is important that the PRACE model become a sustainable European infrastructure, which is not quite the case just yet. The financial commitment for PRACE will run until 2015, that is to say, tomorrow. However, we are building PRACE’s future today, and we will need a different form of financial support, because four countries alone will not be able to support access to high performance computing resources for all the countries of the Union. Other sources of funding will have to be found, whether they come from the European Commission or other partners, to support both the investment in the machines and their operating costs.

At the French level, when GENCI was founded in 2007, the initial budget was €27 million, and the plan was to double it over the next three years. Today it is still €30 million. We are of course fully aware of the difficulties related to the current economic context; they underline the absolute necessity to unite at the European level.

PRACE’s role as a pillar of the development of HPC in Europe being now fully recognized, how is it positioned with respect to the ETP4HPC?

The policy of the European Commission concerning HPC is clear. It is a policy based on three pillars: a pillar of research infrastructure, an R&D pillar to build technologies and a third pillar concerned with applications. ETP4HPC is the technological pillar, PRACE the infrastructure pillar, and the Centers of excellence, still under construction, the applications pillar.

PRACE has just articulated its strategic vision through the year 2020. Could you present its main points and specific objectives for us?

As we just stated, PRACE has guaranteed funding until mid-2015. We have now defined the next phase with the aim of meeting the needs of users. Written by the Scientific Steering Committee on the basis of the Scientific Case that identifies these needs (among others), the PRACE Council agreed to aim for overall computing power on the order of 50 Pflops by 2020, with complementary architectures. We also want to develop different uses for the machines and to facilitate code porting, a practice that did not really exist in the previous PRACE model.

On the other hand, today the Council is working on the financial model for PRACE’s future which, for its part, is still under discussion – and for which there should be greater funding. The objective is to have a clear vision by the end of the year.

Do you have a doubt about the availability of these funds?

Caution is needed, but it’s now clear that PRACE is a success. However, based on the TCO of a 1 Pflops machine, estimated at €100 million over five years, the cost of 50 Pflops of computing power is not of the same order of magnitude. Given these facts, we must come up with a new a new financial model for PRACE.