| Abstract: |
|
Power is a key challenge for exascale computing. We must lay a solid foundation to measure, report and analyze the system power and energy consumed while running a workload. Join panelists Kim Cupps, Lawrence Livermore National Laboratory, Herbert Huber, Leibniz Supercomputing Center, Buddy Bland, Oak Ridge National Laboratory, and Susan Coghlan, Argonne National Laboratory as they review their experiences in taking power measurements while running HPL using a common methodology defined jointly by the Top500, Green500, Green Grid and the Energy Efficient HPC Working Group. These leading organizations are focused on improving performance and energy efficiency of supercomputers and are working together to develop energy efficiency metrics for the world’s most powerful supercomputers. The system architectures that helped break the petascale barrier will not be sufficient for exascale, but by working together to first develop common metrics, we lay the groundwork for innovative approaches needed to significantly improve energy efficiency while continuing to increase system performance. |
|
