The Bavarian Competence Network for Technical and Scientific High Performance Computing: KONWIHR

Computational Science has established itself as the third pillar of scientific enquiry alongside theory and experiment. The discipline of Computational Science enhances scientific investigation by building and testing models of complex phenomena, yielding new information, innovation and fresh insight into the research process that is otherwise not obtainable. Many of these problems require outstanding computational resources going far beyond the capabilities of desktop PCs. Prominent and important examples in this context are the simulation of turbulent flows, the computation of quantum mechanical many-body interactions in theoretical physics, the evolution of protein dynamics in life sciences or the propagation of earthquake waves.

Fostering the use of modern supercomputers within scientifically and industrially oriented research communities to solve scientifically challenging and technically highly relevant problems has been the main focus of the Competence Network for Technical and Scientific High Performance Computing in Bavaria (KONWIHR). KONWIHR has been established by the Federal State of Bavaria in May 2000 with a total budget of about four million EUR has supported 25 to 30 larger scientific projects and numerous short evaluation projects at Bavarian universities in the 2000-2005 timeframe. A special focus of the KONWIHR activities has been on porting, optimization and parallelization of application codes for the Höchstleistungsrechner in Bayern I (HLRB-I) which was the first federal supercomputer at LRZ Munich and, at the time of installation, the first TFlop/s system in Europe.

Projects from the Friedrich-Alexander University Erlangen-Nuremberg have acquired a substantial fraction of the total KONWIHR funding through competitive proposals, reviewed by an external advisory board comprised of experts from universities, research institutes and industry. The thematic foci of the KONWIHR activities in Erlangen have been multi-physics applications in engineering, molecular dynamics simulation in life sciences, parallel methods for highly correlated quantum systems and evaluation of new programming approaches as well as multi-threaded hardware architectures in computer science. Several of the KONWIHR projects document their current activities in this booklet. Additionally, the large variety of contributions which mainly rely on local clusters reflects the successful Bavarian strategy of providing both user support and adequate funding on all levels of the high performance systems hierarchy.

A very important aspect of KONWIHR is that it has paved the way to install highly competitive research and support groups at the computing centres of Erlangen (RRZE) and Munich (LRZ). These groups provide extensive and qualified user training as well as high level user support for all challenges in high performance computing, ranging from code debugging and optimization through massive and hybrid parallelization. Of course, this work is fully integrated into the research activities of the scientific groups, requiring a broad level of scientific HPC research at the computing centres themselves. Within less than a decade an internationally recognised HPC group could be established at RRZE. The HPC expertise of both centres also enabled a smooth transition for many Bavarian research groups from HLRB-I to HLRB-II which was ranked as a TOP 10 computer worldwide in 2007.

The very successful work of KONWIHR has been continued in selected key projects beyond the official lifetime of KOWIHR through follow-on funding provided by the Federal State of Bavaria in 2006 and 2007. A new sustainable infrastructure for HPC in Bavaria is expected to be launched in 2008 through KONWIHR-II. The focus of the new program will be to further extend the services and competence of the HPC centres of excellence at LRZ and RRZE and to provide funding and know-how to leading edge as well as emerging HPC projects. The goal is to prepare user applications for the technological / disruptive changes expected in the area of Petaflop computing such as massive parallel processing (100,000+ cores) or homogeneous and heterogeneous many-core architectures. These activities will be decisive for the future success of Bavaria in the context of the installation of European-level supercomputing centres and the scientific amortization of the related large investments.

The KONWIHR eco system

Back to top