SC13 Denver, CO

The International Conference for High Performance Computing, Networking, Storage and Analysis

Extreme Scaling of Real World Applications to >130,000 Cores on SuperMUC.

Authors: Helmut Satzger (Leibniz Supercomputing Centre), Momme Allalen (Leibniz Supercomputing Centre), Christoph Bernau (Leibniz Supercomputing Centre), Nicolay Hammer (Leibniz Supercomputing Centre), David Brayford (Leibniz Supercomputing Centre), Carmen Navarrete (Leibniz Supercomputing Centre), Ferdinand Jamitzky (Leibniz Supercomputing Centre), Anupam Karmakar (Leibniz Supercomputing Centre), wolfram schmidt (University of Göttingen), Jan Frederik Engels (University of Göttingen), Gurvan Bazin (Universitäts-Sternwarte, Ludwig-Maximilians Universität München), Jürg Diemand (University of Zürich), Klaus Dolag (Universitäts-Sternwarte, Ludwig-Maximilians Universität München), Carsten Kutzner (Max Planck Institute for Biophysical Chemistry, Göttingen), Andreas Marek (Rechenzentrum der Max-Planck-Gesellschaft am Max-Planck-Institut für Plasmaphysik, Garching), Philipp Trisjono (RWTH Aachen), Arndt Bode (Leibniz Supercomputing Centre), Herbert Huber (Leibniz Supercomputing Centre), Matthias Brehm (Leibniz Supercomputing Centre)

Abstract: In July 2013, the Leibniz Supercomputing Centre held the first workshop to test extreme scaling on SuperMUC, the 3 PFLOP/s system with 147,456 Intel Sandy Bridge CPU cores. Groups from 15 international projects came to the LRZ with codes that had could scale up to 4 islands (32,768 cores). During the workshop, the participants tested the scaling capabilities on the whole system. Application experts from the LRZ, Intel and IBM were on site to resolve issues and assist in the performance optimization. New techniques like fast startup were successfully tested which helped to reduce the startup time by a factor of 2-3. At the end of the workshop, 6 applications were successfully running on the full machine, while the other 8 applications managed to run on half of the system. The LRZ is already planning a follow-up workshop where the improvements and feedback from the experts will be tested.

Poster: pdf
Two-page extended abstract: pdf

Poster Index