Dynamic scheduling arithmetics for hard real-time period tasks in Dynamic Voltage Scaling system

Dynamic Voltage Scaling (DVS) has become a promising method for embedded real-time systems to reduce their power consumption. Unlike schedulability, energy consumption focus on the accumulative effect in a period of time. The traditional real-time scheduling model based on the worst execution time is not satisfied for the need of energy consumption analysis. In this paper the real-time task scheduling model is built according to the probability distribution of actual execution time, and the effect of the Actual Execute Time (AET) distribution on the energy consumption performance of real-time DVS scheduling algorithm is analyzed using stochastic method. The real-time DVS dynamic scheduling algorithm based on the average execution time is presented. The experimental results show that algorithm presented in this paper can hold stable power consumption performance under different AET distribution.