Ph.D., Colorado State University, August 2015
Co-Major Professors: H. J. Siegel and Anthony A. Maciejewski
As high performance heterogeneous computing systems continually become faster, the operating cost to run these systems has increased. A significant portion of the operating costs can be attributed to the amount of energy required for these systems to operate. To reduce these costs it is important for system administrators to operate these systems in an energy efficient manner. Additionally, it is important to be able to measure the performance of a given system so that the impacts of operating at different levels of energy efficiency can be analyzed. The goal of this research is to examine how energy and system performance interact with each other for a variety of environments. One part of this study considers a computing system and its corresponding workload based on the expectations for future environments of Department of Energy and Department of Defense interest. Numerous heuristics are presented that maximize a performance metric created using utility functions. Additional heuristics and energy filtering techniques have been designed for a computing system that has the goal of maximizing the total utility earned while being subject to an energy constraint. A framework has been established to analyze the trade-offs between performance (utility earned) and energy consumption. Stochastic models are used to create “fuzzy” Pareto fronts to analyze the variability of solutions along the Pareto front when uncertainties in execution time and power consumption are present within a system. In addition to using utility earned as a measure of system performance, system makespan has also been studied. Finally, a framework has been developed that enables the investigation of the effects of P-states and memory interference on energy consumption and system performance.