Uncore Power Scavenger: A Runtime for Uncore Power Conservation on HPC Systems
Event Type
Registration Categories
Deep Learning
Extreme Scale Computing
Machine Learning
Resource Management
Runtime Systems
TimeTuesday, 19 November 20193:30pm - 4pm
DescriptionThe US Department of Energy (DOE) has set a power target of 20-30MW on the first exascale machines. To achieve one exaflop under this power constraint, it is necessary to minimize wasteful consumption of power while striving to improve performance. Toward this end, we investigate uncore frequency scaling (UFS) as a knob for reducing power footprints of HPC jobs. We propose Uncore Power Scavenger (UPSCavenger), a runtime system that dynamically detects phase changes and automatically sets the best uncore frequency for every phase to save power without significant impact on performance. Our experimental evaluations show that UPSCavenger achieves up to 10% energy savings with under 1% slowdown. It achieves 14% energy savings with the worst case slowdown of 5.5%. We show that UPSCavenger achieves up to 20% speedup and proportional energy savings compared to Intel’s RAPL with equivalent power usage making it a viable solution even for power-constrained computing.
Back To Top Button