Frontier: Using Real-Time Power Monitoring to Assess System Health and Performance

In this talk, Scott Atchley will discuss the challenges faced by the Oak Ridge Leadership Computing Facility (OLCF) as they tried to achieve an exaflop HPL result using the Frontier exascale supercomputer. He will describe how OLCF, HPE, and AMD used real-time power monitoring as an aid to understand system health as well as a benchmark for assessing system performance. He will also discuss Frontier's design to maximize cooling and power efficiency that allowed Frontier and its single-cabinet Test and Development System (TDS) to take the top two spots on the Green500 at last year's ISC.