An Energy-Efficient Scheduling Algorithm for Shared Facility Supercomputer Centers

The evolution of high-performance computing is associated with the growth energy consumption. Performance cluster computes (is increased via rising in performance and number used processors, GPUs, coprocessors. An increment elements results significant Power grids limits for supercomputer centers (SCC) are driving transition to more energy-efficient solutions. Often upgrade resources done step-by-step, i.e. parts older supercomputers removed from service replaced newer ones. A single SCC at any time can operate several systems different power That why problem scheduling parallel programs execution on optimize consumption minimize increase (energy-efficient scheduling) important. goal presented work was development a new algorithm SCC. To reach authors analyzed methods shared facility, including minimizing methods. study made it possible formulate set CCs propose an its solution. Experiments NPB benchmarks allowed achieving reduction minor runtime.