Integrating Preemption Threshold Scheduling and Dynamic Voltage Scaling for Energy Efficient Real-Time Systems

Preemption threshold scheduling (PTS) enables designing scalable real-time systems. PTS not only decreases the run-time overhead of the system, but can also be used to decrease the number of threads and the memory requirements of the system. In this paper, we combine preemption threshold scheduling with dynamic voltage scaling to enable energy efficient scheduling in real-time systems. We consider scheduling with task priorities defined by the Earliest Deadline First (EDF) policy. We present an algorithm to compute threshold preemption levels for tasks with given static slowdown factors. The proposed algorithm improves upon known algorithms in terms of time complexity. Experimental results show that preemption threshold scheduling reduces on an average 90% context switches, even in the presence of task slowdown. Further, we describe a dynamic slack reclamation technique that working in conjunction with PTS that yields on an average 10% additional energy savings.