High Performance Computing in Science and Engineering ’06

Abstract

The last two years have been great for high performance computing in Baden- W¨ urttemberg and beyond. In July 2005, the new building for HLRS as well as Stuttgart’s new NEC supercomputer – which is still leading edge in Ger- many – have been inaugurated. In these days, the SSC Karlsruhe is finalizing the installation of a very large high performance system complex from HP, built from hundreds of Intel Itanium processors and more than three thou- sand AMD Opteron cores. Additionally, the fast network connection – with a bandwidth of 40 Gbit/s and thus one of the first installations of this kind in Germany – brings the machine rooms of HLRS and SSC Karlsruhe very close together. With the investment of more than 60 Million Euro, we – as the users of such a valuable infrastructure – are not only thankful to science managers and politicians, but also to the people running these components as part of their daily business, on a 24-7 level. Since about 18 months, there are lots of activities on all scientific, advisory, and political levels to decide if Germany will install an even larger European supercomputer, where the hardware costs alone will be around 200 Million Euro for a five year period. There are many good reasons to invest in such a program because – beyond the infrastructure – such a scientific research tool will attract the best brains to tackle the problems related to the software and methodology challenges. Within the last six months, under the guidance of Professor Andreas Reuter (EML) the German HPC community has made a proposal to reshape High Performance Computing in Germany and to form a German HPC Alliance, with the goal to improve and guarantee competi- tiveness for the coming years. There is a very good chance that our Federal Ministry for Education and Research (BMBF) – together with the colleagues from the local countries – will support the proposed actions. Beyond the stabi- lization and strengthening of the existing German infrastructures – including the necessary hardware at a worldwide competitive level – a major request is a related software research and support program to enable Computational Sci- ence and Engineering on the required level of expertise and performance which means: running Petascale applications on more than 100,000 processors. It is recommended that for the next years 20 Million Euro are spend – on a yearly basis – for projects to develop algorithms, methods and tools in many areas. As we all know, we do not only need competitive hardware but also excellent software andmethods to approach – and solve – the most demanding problems in science and engineering. To achieve this challenging goal every three-year project supported by that program will need to integrate excellent research groups at the universities with colleagues from the competence network of HPC centers in Germany. The success of this approach is of utmost impor- tance for our community and also will strongly influence the development of new technologies and industrial products; beyond that, this will finally deter- mine if Germany will be an accepted partner among the leading technology and research nations. Since 1996, HLRS is supporting the scientific community as part of its of- ficial mission. Like in the years before, the major results of the last 12 months were reported at the Tenth Results and Review Workshop on High Perfor- mance Computing in Science and Engineering, which was held October 19–20, 2006 at Stuttgart University. This volume contains the written versions of the research work presented. The papers have been selected from all projects run- ning at HLRS and at SSC Karlsruhe during the one year period beginning October 2005. Overall, 39 papers have been chosen from Physics, Solid State Physics, Computational Fluid Dynamics, Chemistry, and other topics. The largest number of contributions, as in many other years, came from CFD with 17 papers. To a certain extend, the selected papers demonstrate the state of the art in high performance computing in Germany. The authors were encouraged to emphasize computational techniques used in solving the problems examined. The importance of the newly computed results for the specific disciplines, as interesting as they may be from the scientific point of view, were not the major focus of this volume. We gratefully acknowledge the continued support of the Land Baden- W¨ urttemberg in promoting and supporting high performance computing. Grateful acknowledgement is also due to the Deutsche Forschungsgemein- schaft (DFG): many projects processed on the machines of HLRS and SSC could not have been carried out without the support of the DFG. Also, we thank the Springer Verlag for publishing this volume and thus helping to po- sition the local activities into an international frame.We hope that this series of publications is contributing to the global promotion of high performance scientific computing. Stuttgart,