Presentation: Electric Grid and High Performance Computing (HPC) Challenges

The High Performance Computing center at Lawrence Livermore National Lab is a large Megawatt user facility. Sequoia, one of the HPC machines and ranked number 3 on the top 500, is rated at 9.6MW peak with 20 petaflops. The workload is burst with power swings that are over 4 MW. The HPC center at LLNL is concerned about the quality, cost, environmental impact, and availability of electricity, our providers are concerned about electrical grid reliability, particularly in terms of energy consumption, peak power demands, and power fluctuations. These demands led to the development of new relationships. These relationships are driven by a mutual interest to reduce energy costs and improve electrical grid reliability.

Improvement cannot be made if there is no data or measurement. LLNL implemented an extensive monitoring and data collection system to collect, analyze, and correlate data from legacy power meters, PMU's, and computer environmental and power sensors. HPC uses the data to understand and enhance power usage, track and analyze abnormal events, and improve energy efficiency. While calculating energy efficiency metric for the High Performance Computing facility, LLNL discovered a problem with a meter on the transformer. The meter was reporting incorrect values and is being fixed to ensure data accuracy. LLNL also uses the PI System data to characterize the Sequoia power profile and respond to requests from their energy provider to provide an hourly power usage schedule 24 hours ahead of time. The data is also used to calculate two power efficiency metrics (PUE and TUE) to help increase energy efficiency and minimize energy consumption."