CPU scheduling for statistically-assured real-time performance and improved energy efficiency

We present a CPU scheduling algorithm, called energy-efficient utility accrual algorithm (or EUA), for battery-powered, embedded real-time systems. We consider an embedded software application model where repeatedly occurring application activities are subject to deadline constraints specified using step time/utility functions. For battery-powered embedded systems, system-level energy consumption is also a primary concern. We consider CPU scheduling that (1) provides assurances on individual and collective application timeliness behaviors and (2) maximizes system-level timeliness and energy efficiency. Since the scheduling problem is intractable, EUA heuristically computes CPU schedules with a polynomial-time cost. Several properties of EUA are analytically established, including timeliness optimality during under-load situations and statistical assurances on timeliness behavior. Further, our simulation results confirm EUA's superior performance.