EHS: Energy-Aware CPU and I/O Scheduling for Embedded, Real-Time Systems

点击翻译按钮获取中文摘要

CNS 0409382 Stephen M. Goddard University of Nebraska-LincolnTitle: Energy-Aware CPU and I/O Scheduling for Embedded, Real-Time SystemsEmbedded real-time systems often have severe power constraints that require aggressive energy conservation techniques in both the processor and the Input/Output (I/O) devices. This project is investigating energy-aware processor and I/O scheduling for embedded, hard real-time systems with flexible task models. The results will provide the opportunity for increased energy savings in reactive, embedded, real-time systems that execute in unpredictable environments, such as mobile robots. The project focuses on two central problems: processor energy conservation for sporadic tasks, and energy conservation for I/O devices with periodic tasks.Processor Energy Conservation: Processor energy conservation for the periodic task model has been extensively researched in the context of dynamic power consumption for embedded, real-time systems. However, energy conservation with more flexible real-time task models, such as the sporadic task model, remains an open problem. The project is investigating processor energy conservation for sporadic tasks that have deadlines not equal to their respective periods, non-preemptive earliest-deadline-first (EDF) scheduling, shared resources under EDF scheduling, and generalized versions of the sporadic task model. Energy Conservation in I/O Devices with Periodic Tasks: This research is investigating methods that dynamically re-order job executions to conserve energy in I/O devices under the periodic task model. Though the focus of this aspect of the project is energy conservation in I/O devices, the method minimizes processor dynamic power dissipation when only frequency scaling is supported (and not voltage scaling). It also reduces leakage power consumption by grouping short idle periods into longer intervals of time in which the processor can be made idle without jeopardizing temporal correctness.