Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen (Leibniz Universität Hannover) |
The High Performance Computing Center at the Gottfried Wilhelm Leibniz University (LUH-HLRN) hosts the supercomputer of the North-German Supercomputing Alliance, HLRN. This Alliance was founded in 2001 and the first supercomputer (an IBM-system) at HLRN was installed in 2002. Currently, the second generation high performance computer HLRN-II, a SGI Altix Cluster of approx. 300 TFlops peak performance (in total about 25.000 cores), is in operation and used by more than 140 research projects from all over Northern Germany. The High Performance Computing Center at Leibniz University is led by PD Dr. Steffen Schulze-Kremer since December 2003.
The procurement process for the third generation, 30 million Euro supercomputer concluded successfully in December 2012. The new HLRN-III will be a Cray Cascade System and will be installed, like the previous two generations, in two equal parts at two sites: one at the High Performance Computing Center at the Gottfried Wilhelm Leibniz University and the other at the Zuse-Institute Berlin. Both systems are connected via a high speed data link and can be operated as one system. Operation of HLRN-III is planned to commence before the end of 2013.
Prospective users of the HLRN-System submit research proposals to the scientific board of the HRLN community. If judged to be of sufficient scientific merit based on DFG criteria, successful applications are granted access to the computing and data resources at HLRN. For NUMEXAS HLRN will contribute dedicated access to the HLRN facilities to the project consortium together with scientific support to optimize applications in this environment.
The whole HLRN team is distributed over several sites, since several federal states in Northern Germany have a share in HLRN. The LUH-HLRN team (6 scientists) has more than 10 years experience in HPC computer operation, user support including optimization of user programs and training users in parallel computing.
Steffen Schulze-Kremer has a Diploma degree in Biochemistry from the Free University of Berlin (FUB), a PhD in Biochemistry (FUB) and a Habilitation (http://en.wikipedia.org/wiki/Habilitation) in Bioinformatics (FUB). Since 2003 he leads the High Performance Computing Center at Leibniz University Hannover (LUH-HLRN).
Gerd Brand has a Diploma degree in Mathematics and a PhD in Applied Mathematics from the Leibniz University Hannover (LUH). Currently he is working at the LUH site of the HLRN as a HPC Support Scientist.
Holger Naundorf has a Diploma degree in Physics and a PhD in Theoretical Chemistry. Currently he is working at the LUH site of the HLRN as a HPC Support Scientist.
LUH-HLRN will provide up-to-date HPC-hardware (Cray Cascade system, part of the system with accelerators) and actual software tools for program development and optimization to be used on this system. Furthermore tools for power measurement for the provided platform will be available to determine the energy consumption of simulation runs.
On the application software side initially (parallel) numerical libraries will be installed and evaluated. At first the focus will be on Trilinos, since this library is used by the software packages and has an interface for manycore compute nodes including some accelerators (Trilinos package Tpetra). As the case may be other libraries have to be examined, for example PETSc or vendor specific libraries. Secondly the software of WPs 3-7 will be treated. This means installation, testing, measurement of performance and energy consumption and taking scaling experiments on the HLRN-III platform. Steps in optimization will be carried out in close collaboration with the software developers of WPs 3-7. WPs affected: 3, 8
The research leading to these results has received funding from the European Community's Seventh Framework Programme under grant agreement n° 611636