- The new website can be found at https://eehpcwg.llnl.gov/ -
(please update your bookmarks)
Liquid cooling is key to dealing with heat density, energy efficiency, and increasing the performance of supercomputers. It becomes even more predominant looking forward. The transition to liquid cooling, however, comes with challenges.
Each system and each site comes with its specific issues regarding liquid cooling. These can complicate procurements, design, installation, operations, and maintenance.
This BoF will review the immediate past history from a lessons learned perspective as well as discuss what’s needed for liquid cooling to be implemented more readily in the future. It will explore HPC center’s liquid cooling design parameters, commissioning guidelines and control systems.
US legislation just passed requiring energy efficiency metrics for federal data centers. Without heavy lifting, the metric will be PUE. However, with PUEs approaching 1.1, PUE may have limited utility.
Should and can we move on from PUE? This BoF discusses what’s needed for driving the next level of improvement. It will also discuss advancing system-level energy efficiency workloads and metrics from the currently-used HPL FLOPS/Watt.
The BoF will host a panel and field questions from the floor. Panelists will include experts in the field such as Erich Strohmaier, Top500 & LBNL and Michael Patterson, The Green Grid & Intel.
Supercomputing centers are poised to begin a transition to “dynamic power management.” Multiple forces are driving this need with energy cost control in the forefront. Supercomputer systems have increasingly rapid, unpredictable and large power fluctuations with costly implications for the supercomputer center. In addition, electricity service providers may incentivize supercomputing centers to change their timing and/or magnitude of demand and to help address variable renewable generation. To adapt to this new landscape, supercomputing centers may employ strategies to dynamically and in real-time control their electricity demand. This BoF seeks those interested in sharing experiences with dynamic power and energy management.
The Green500, entering its eighth year, encourages sustainable supercomputing by raising awareness in the energy efficiency of such systems. This BoF will present (1) evolving metrics, methodologies, and workloads for energy-efficient HPC, (2) trends across the Green500, including the trajectory towards exascale, and (3) highlights from the latest Green500 List. In addition, the BoF will discuss a collaborative effort between Green500, Top500, and EE HPC WG to improve power measurement while running a workload, such as HPL, and solicit feedback from the HPC community. The BoF will close with an awards presentation, recognizing the most energy-efficient supercomputers in the world.