A three step blind approach for improving high performance computing systems’ energy performance

摘要

Nowadays, there is no doubt that energy consumption has become a limiting factor in the design and operationrnof high performance computing (HPC) systems. This is evidenced by the rise of efforts both fromrnthe academia and the industry to reduce the energy consumption of those systems. Unlike hardware solutions,rnsoftware initiatives targeting HPC systems’ energy consumption reduction despite their effectivenessrnare often limited for reasons including (1) the program specific nature of the solution proposed; (2) thernneed of deep understanding of the application task at hand; and (3) proposed solutions are often difficultrnto use by novices and/or are designed for single task environments. This paper proposes a three step blindrnsystem-wide, application independent, fine-grain, and easy to use (user friendly) methodology for improvingrnenergy performance of HPC systems. The methodology typically breaks into phase detection, phase characterization,rnand phase identification and system reconfiguration. And it is blind in the sense that it doesrnnot require any knowledge from users. It relies upon reconfigurable capabilities offered by the majority ofrnHPC subsystems—including the processor, storage, memory, and communication subsystems—to reducernthe overall energy consumption of the system (excluding network equipments) at runtime. We also presentrnan implementation of our methodology through which we demonstrate its effectiveness via static analysesrnand experiments using benchmarks representative of HPC workloads. The memory becoming one ofrnthe most power hungry HPC system this paper also introduces and investigates the relevance of a potentialrnpower saving scheme which we refer to as memory size scaling. It is destined to scale the memory size forrnsaving energy in a CPU frequency scaling like fashion.