The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Authors: Wu Feng (Virginia Tech), Erich Strohmaier (TOP500), Eric Whiting (Idaho National Laboratory), Natalie Bates (Energy Efficient HPC Working Group), Matthias Maiterth (Oak Ridge National Laboratory (ORNL))
Abstract: With power being a first-order design constraint on par with performance, it is important to measure and analyze energy-efficiency trends in supercomputing. To raise the awareness of greenness as a first-order design constraint, the Green500 seeks to characterize the energy efficiency of supercomputers for different metrics, workloads, and methodologies. This BoF discusses trends across the Green500 and highlights from the current Green500 list. In addition, the Green500, Top500, and Energy Efficient HPC Working Group have been working together on improving power-measurement methodology, and this BoF presents recommendations for changes that will improve ease of submission without compromising accuracy.
Long Description: To raise the awareness of greenness as a first-order design constraint, the Green500 seeks to characterize the energy efficiency of supercomputers. Based on this characterization, it also provides a ranking of the most energy-efficient supercomputers in the world.
The goals of this BoF are multi-fold: (1) analyze and announce the latest Green500 List, (2) present technical details and analysis on three of the most energy-efficient supercomputers, (3) and present analysis on improving quality and accuracy aspects of the power measurement methodology.
This joint BoF between the Green500, Top500, and the Energy Efficient HPC Working Group (EE HPC WG) has been held every year at SC since SC11. The participation has varied between 75 and 125 people.
Feedback on the power measurement methodology more generally will be captured. This information will be used to make improvements to the methodology. Currently, the methodology has three levels of quality. Level 3 (L3) captures the energy and power of the entire system and is the most accurate measurement. Level 2 (L2) and level 1 (L1) captures the power used for a fraction of the system, with L2 more accurate than L1.
At the SC21 BoF, we announced the creation of an open community-based team [ https://sites.google.com/view/eehpcwg-methodology/home ] chartered to review and make proposed improvements to the Energy Efficient High Performance Computing Power Measurement Methodology. This is the methodology that is used for power submissions to the Top500 and Green500. The technical lead is Thomas Ilsche and there is active participation from ~50 people from other HPC sites, the vendor community and academe. Since 2021, this team has continued to seek feedback from the wider community. Most recently, the team has been soliciting feedback from sites that made an HPL submission to the Top500, but either did not make a Green500 power submission at all, or made a power submission without any indication of the methodology used. This team will be reviewing this feedback at the SC25 BoF.
The Green500 List will be published on the Green500 website. All other presentations (technical details on the three most energy efficient supercomputers as well as case studies on power measurement methodology) will also be published on the EE HPC WG website.