System-level power-aware scheduling by operation-based prediction

Mobile devices have become more and more popular, and portable systems require long battery lifetime to provide services as long as possible while still delivering high performance. Dynamic power management (DPM) policies trade the performance for the power saving by selectively placing components into low-power states. In this paper, we propose a power-aware OS scheduler for reducing power consumption. We also extend the battery lifetime by predicting the IO-burst in each process. In round robin scheduling, each process is executed into several rounds, and the IO-burst can be predicted precisely. When processes are properly scheduled, we can save power consumption without losing performance. We also consider battery characteristics to schedule processes to make the capacity of the battery to be approximate to ideal. Experimental results show that our method achieves high hit rate and the battery lifetime can be extended to 38%.