Real-Time Scheduling of Tasks with Energy Constraints
Abstract
Energy consumption is a significant factor in the design of digital systems. The importance is even more notorious when contemplating portable and embedded systems, due to its operational dependancy of batteries. As processors are a preeminent source of energy consumption, novel techniques that fully exploit its hardware characteristics permit the improvement of energy gains. However, the costly trade-off is represented by the decrease of speed and consequent performance degredation. In this context, real-time systems provide the timeliness guarantees that allow to jointly consider energy reduction and minimum performance requirements. In this thesis, we address two scheduling policies, Round-Robin and Dual-Priority, for energy minimization purposes. Considering Round-Robin and Dual-Priority, the optimal discrete processor speed for scheduling job sets is discovered, founded in an efficient feasibility test, that is also proposed. Regarding Dual-Priority, the study of the mixed workload of soft and hard real-time tasks provided and opposing constraint of the energy minimization problem: the minimization of response times. Based in this ambivalence of the system, a design metric is defined. A derived importance concept of this metric is used in a methodology that supports engineering decisions in early stages of the real-time system design. A revised worst case response times formula for FPP is also presented, when tasks are scheduled with different processor speeds.