The Impact of Energy-Efficient Scheduler Overhead on the Performance of Embedded Real-Time Systems

In this paper, we investigate the impact of scheduler overhead on energy-efficient, real-time scheduling algorithms for battery-powered, mobile embedded systems. We consider algorithms based on deadlines, and those based on time/utility functions (TUFs) and utility accrual (UA) scheduling criteria. The complete experiments include implementation of prototype algorithms in a Linux kernel, and measurement of the actual energy consumption and system performance. We plan to analyze the factors that can possibly affect the system-level performance with scheduler overhead, such as system load, energy model settings, and TUFs, and identify the conditions under which the TUF/UA algorithms have superior performance, despite their high overhead. The preliminary experiments in this paper consider important parameters, such as task execution times, for our future kernel implementation and measurements.