Date of Award
Doctor of Philosophy (PhD)
Electrical Engineering and Computer Science
Power consumption has become a major concern in the design of computing systems today. High power consumption increases cooling cost, degrades the system reliability and also reduces the battery life in portable devices. Modern computing/communication devices support multiple power modes which enable power and performance tradeoff. Dynamic power management (DPM), dynamic voltage and frequency scaling (DVFS), and dynamic task migration for workload consolidation are system level power reduction techniques widely used during runtime. In the first part of the dissertation, we concentrate on the dynamic power management of the personal computer and server platform where the DPM, DVFS and task migrations techniques are proved to be highly effective. A hierarchical energy management framework is assumed, where task migration is applied at the upper level to improve server utilization and energy efficiency, and DPM/DVFS is applied at the lower level to manage the power mode of individual processor. This work focuses on estimating the performance impact of workload consolidation and searching for optimal DPM/DVFS that adapts to the changing workload. Machine learning based modeling and reinforcement learning based policy optimization techniques are investigated.
Mobile computing has been weaved into everyday lives to a great extend in recent years. Compared to traditional personal computer and server environment, the mobile computing environment is obviously more context-rich and the usage of mobile computing device is clearly imprinted with user's personal signature. The ability to learn such signature enables immense potential in workload prediction and energy or battery life management. In the second part of the dissertation, we present two mobile device power management techniques which take advantage of the context-rich characteristics of mobile platform and make adaptive energy management decisions based on different user behavior. We firstly investigate the user battery usage behavior modeling and apply the model directly for battery energy management. The first technique aims at maximizing the quality of service (QoS) while keeping the risk of battery depletion below a given threshold. The second technique is an user-aware streaming strategies for energy efficient smartphone video playback applications (e.g. YouTube) that minimizes the sleep and wake penalty of cellular module and at the same time avoid the energy waste from excessive downloading.
Runtime power and thermal management has attracted substantial interests in multi-core distributed embedded systems. Fast performance evaluation is an essential step in the research of distributed power and thermal management. In last part of the dissertation, we present an FPGA based emulator of multi-core distributed embedded system designed to support the research in runtime power/thermal management. Hardware and software supports are provided to carry out basic power/thermal management actions including inter-core or inter-FPGA communications, runtime temperature monitoring and dynamic frequency scaling.
Shen, Hao, "ADAPTIVE POWER MANAGEMENT FOR COMPUTERS AND MOBILE DEVICES" (2014). Dissertations - ALL. 197.