The 7th Annual Meeting on High Performance Computing and Infrastructure in Norway

NOTUR2008 - ABSTRACTS

Towards Petaflops computing with JUGENE

Norbert Attig

Driven by technology, Scientific Computing is rapidly entering the PetaFlops era. The Jülich Supercomputing Centre (JSC), one of three German national supercomputing centres, is focusing on the IBM Blue Gene architecture to provide computer resources of this class to its users. Following the successful operation of a large Blue Gene/L system at JSC in the years 2006/7 a 16-rack Blue Gene/P with 65,536 processors was installed in October 2007. With a peak performance of 222.8 TFlop/s, Jülich's Blue Gene/P - alias JUGENE - is currently the biggest supercomputer in Europe and ranked No 2 worldwide. A key feature of this architecture is its scalability towards PetaFlops computing based on low power consumption, small footprint and an outstanding price-performance ratio.

From the user perspective it is of highest importance that many applications from different research fields can be ported to, scale well and run efficiently on this new architecture. Blue Gene applications at JSC cover a broad spectrum ranging from LQCD to MD codes like CPMD and VASP, materials science, protein folding codes, fluid flow research, quantum computing and many, many others. However, JUGENE is used as a Leadership-class system and hosts only a small number of projects to give selected researchers the opportunity to get new insights into complex problems which were out of reach before. The performance and the scaling behaviour of the applications are being continuously improved in close collaboration between the user support team at JSC and the corresponding computational scientists.

For example, JSC organizes Blue Gene Scaling Workshops, where experts from the Blue Gene Consortium, comprising Argonne National Laboratory, IBM and Jülich, help to further optimise important applications.

Computational scientists from many research areas use this chance to improve their codes during these events and then later apply for significant shares of Blue Gene computer time to tackle challenging phenomena which come into reach with this architecture.

< Back to programme