| In this dissertation, we model power, energy, and temperature across a large range of systems from the device level to the cluster level. By understanding how software, hardware, and work-loads each affect power, energy, and temperature, we propose techniques for managing these factors under a variety of different scenarios.; In the first study, we examine application transformations that increase device idle times and inform the operating system about the length of each upcoming idle time. We use modeling and experimentation to assess the potential energy and performance benefits of this type of application support for a laptop disk. Overall, we find that the transformations can reduce disk energy consumption by as much as 85% with only a small degradation in performance.; In the second study, we argue that designing efficient servers for heterogeneous clusters requires defining an efficiency metric, modeling the different types of nodes with respect to the metric, and searching for request distributions that optimize the metric. To concretely illustrate this process, we design a cooperative Web server for a heterogeneous cluster that uses modeling and optimization to minimize the energy consumed per request. Our experimental results for a cluster comprised of traditional and blade nodes show that our server can consume 42% less energy than an energy-oblivious server, with only a negligible loss in throughput. The results also show that our server conserves 45% more energy than an energy-conscious server that was previously proposed for homogeneous clusters.; In the third and final study we introduce Mercury, a software suite that emulates system and component temperatures based on simple layout, hardware, and component-utilization data. This tool suite is designed to assist systems researchers working on software-based thermal management, by removing the need for real temperature experiments requiring an isolated and heavily instrumented environment; commercial temperature simulators which may be expensive, hard to use, slow for realistic systems, and do not execute applications or systems software; or circuit and architecture-level temperature simulators which only model the processor and do not execute the systems software. We validate Mercury using real measurements as well as a widely used commercial simulator. We evaluate Mercury by using it to develop Freon, a system that manages component temperatures in a server cluster without unnecessary performance degradation. |
