SuperMUC receives award at ISC'14
8 July 2014Tweet
Researchers using SuperMUC at the LeibnizSupercomputing Centre in Munich, Germany have won the PRACE-ISC Award 2014, presented at this year’s ISC'14 Conference in Leipzig, Germany.
The award, for the highest sustained petascale performance, was presented to a team of scientists for their seismic science project ‘SeisSol’. Using all 147,456 compute cores of the HPC system SuperMUC, the research team achieved a sustained system performance of 1.09 Petaflops, running the application for more than 3 hours to simulate the Earth’s vibrations inside the Merapi volcano on Java island, Indonesia.
Supported by staff from the Leibniz Supercomputing Centre (LRZ), computer scientists, mathematicians, and geophysicists from the Technische Universität München (TUM) and the Ludwig-Maximilians-Universität München (LMU) collectively optimised the code. This was an enormous task as the 70,000 lines of code that make up SeisSol had to be optimised to demonstrate the power of the parallel architecture of SuperMUC.
The collaborative effort, under leadership of Professor Dr Michael Bader (TUM) and Dr Christian Pelties of the Department of Geo and Environmental Sciences at LMU, resulted in achieving a SeisSol application performance of 1.42 Petaflops for a weak scaling test which corresponds to 44.5 per cent of SuperMUC’s peak processing performance. For the entire simulation run a sustained system performance of 1.09 Petaflops was achieved.
‘We cannot stress enough how collaborative efforts of scientists of various fields can lead to substantial success’, said Professor Arndt Bode, Director of the Leibniz Supercomputing Centre. ‘The sustained Petaflops performance obtained in this earthquake simulation could only be achieved through the optimal cooperation of specialists from geophysics, computer science, and from the supercomputing centre as put into practise at LRZ’s Extreme Scaling workshops and through LRZ’s partnership initiatives,’ said Bode.
The basis for the optimisation of this simulation was developed at an Extreme Scaling Workshop held earlier at LRZ, where programmers of the most advanced applications running on SuperMUC together with LRZ’s staff jointly worked on optimising and further parallelising the codes.
At a second Extreme Scaling Workshop, twelve international participants were invited to work for an entire week at LRZ to get their eight application programs efficiently utilising all 147,456 compute cores of SuperMUC in parallel. This resulted in six of the eight application codes being developed sufficiently so that they could run on the full SuperMUC system.