Prof. Dr. Dr. Thomas Lippert, Director of the Institute for Advanced Simulation, Head of Jülich Supercomputing Centre.

Keynote

Knocking at Petaflop’s Door

Just eleven years ago, the first supercomputer, Intel’s ASCI Red, broke the barrier of more than one trillion (short-scale number system) floating point operations per second: one Teraflop/s. Last month, a new system at the Los Alamos National Lab, nicknamed ‘Roadrunner’, achieved for the first time the staggering result of one Petaflop/s or one quadrillion floating point operations per second. Roadrunner was benchmarked through the Linpack benchmark code, which is used to compare and rank the performance of the 500 fastest supercomputers worldwide (http://www.top500.org). The enormous 1000-fold increase in performance of supercomputers in the last eleven years highlights the dramatic progress of these universal instruments. Supercomputers are the base on which has been built the continuous advancement of computer simulations, now a key element in the scientific and industrial competitiveness of knowledge-based economies in the 21st century. Simulations are the engines for industrial fields such as aeronautics, and the automotive, pharmaceutical, oil and financial industries. They drive progress in key scientific fields of the highest societal relevance, like climatology, fusion energy and biology, not to mention the defence sector with its requirement for reliability testing and maintenance of nuclear weapons without the use of nuclear testing.

It is not by chance that it is an American system that has made the enormous leap over the Petaflop/s mark: the USA recognized the growing relevance of simulation science at an early stage. In 1991, they passed the High-Performance Computing Act, which states that HPC be given top priority for research. This initiative had a very strong impact on the coordination of all programs with dedicated budgets for supercomputing, and secured the US’s leadership in the field. Since then, more than half of the fastest systems worldwide have been located in the US. The importance of HPC has also penetrated all levels of the federal administration, and various presidents have declared their strong engagement in this sector on a number of occasions.

Parallel to the USA, which started the race to Petaflop/s around 1997, Japan has also pursued a very active policy in support of HPC. Developed and integrated by NEC in Japan, the ‘Earth Simulator’ was the most powerful supercomputer in the world from 2002 to 2004. Japan’s next-generation supercomputer project aims at delivering 10 Petaflop/s in the year 2012. In addition, China recently announced that it will design and build new supercomputers and join the leading countries in 2010 with the installation of a top-level system.

Europe’s decision makers have so far placed supercomputing for simulations in science and engineering on a much lower level of priority. Europe’s previous framework programme, FP6, concentrated its efforts mostly on embedded systems, telecommunications, distributed computing and data services.

Fortunately there is good reason to be optimistic that the simulation sciences – often denoted as the third column of knowledge creation – will also flourish over here. Support for supercomputing infrastructures has been established in the 7th Framework Programme, and following the late 2007 recommendations of the European Strategy Forum on Research Infrastructures, the European Commission has put the creation of a European supercomputing infrastructure on its agenda. These top-level systems will provide levels of performance comparable to other top installations worldwide for the benefit of Europe’s scientific and technological development.

In spring 2007, fourteen European countries formed the initiative ‘Partnership for Advanced Computing in Europe’ (PRACE), in order to be able to respond to the Commission’s call for proposals. The PRACE project was granted by the European Commission in the autumn of 2007 and commenced in January 2008 (see http://www.prace-project.eu). It is coordinated by the German Forschungszentrum Jülich. Its primary objective is the creation of an organizational and legal framework for a European supercomputing infrastructure by 2010. The new ‘tier-0’ systems (five European countries – the principal partners France, Germany, Spain, the Netherlands and the United Kingdom – announced their support for the infrastructure through the installation of tier-0 systems) will establish a new level of performance in Europe, built of the national HPC infrastructures and the European Grids.

The second important mission of PRACE is to create in Europe a framework for the development of the next generation of supercomputers and revitalize the European supercomputer industry. Leading HPC companies in Europe are keen to contribute, and together with research institutes and universities have formed the consortium PROSPECT in order to cooperate with PRACE. Their vision is to create a European technology platform for supercomputing.

The first signs are now visible of a resurgence of European systems in the top500 list. With the help of PRACE, Europe will soon be knocking at Petaflop’s door, and Europe’s industry may in fact become strong enough to pave the way towards Exaflop performance.

Thomas Lippert