Energy-aware Scheduling for Multiprocessor Real-time Systems

Modern real-time applications have become more sophisticated and complex in their behaviour over the time. Contemporaneously, multiprocessor architectures have emerged. Multiprocessor systems, due to their autonomy and reliability, face critical problem of energy consumption. To address this issue in real-time systems, many software-based approaches have emerged. This thesis proposes new techniques for energy-efficient scheduling of multiprocessor systems. Our first contribution is a hierarchical scheduling algorithm that allows restricted migration of tasks. This algorithm aims at reducing the sub-optimality of global EDF algorithm. The second contribution of this thesis is a dynamic power management technique called Assertive Dynamic Power Management (AsDPM). This technique is an admission control technique for real-time tasks, which decides when exactly a ready task shall execute, thereby reducing the number of active processors. The third contribution of this dissertation is a DVFS technique, referred as Deterministic Strech-to-Fit (DSF) technique, which falls in the category of inter-task DVFS techniques. Both DPM and DVFS techniques are efficient for specific operating conditions. However, they often outperform each other when these conditions change. Our fourth and final contribution is a generic power / energy management scheme, called Hybrid Power Management (HyPowMan) scheme. This scheme, instead of designing new power / energy management techniques for specific operating conditions, takes a set of well-known existing policies. At runtime, the best-performing policy for given workload is adapted by HyPowMan scheme through machine-learning approach.