For the past decades, we have experienced an aggressive technology scaling due to the tremendous advancements of Integrated Circuit technology. As massive integration continues, the power consumption of the IC chips exponentially increases which further degraded the system reliability. This in turn poses significant challenges to the design of real-time autonomous systems. In this paper, we target the problem of designing advanced real-time scheduling algorithms that are subject to timing, energy consumption and fault-tolerant design constraints. To this end, we first investigated the problem of developing scheduling techniques for uniprocessor real-time systems that minimizes energy consumption while still tolerating up to k transient faults to preserve the system's reliability. Two scheduling algorithms are proposed: the first scheduler is an extension of an optimal fault-free energy-efficient scheduling algorithm, named ES-DVFS. The second algorithm aims to enhance the energy saving by reserving adequate slack time for recovery when faults strike. We derive a necessary and sufficient condition that must be efficiently checked for the time and energy feasibility of aperiodic jobs in the presence of failures. Later, we formally prove that the proposed algorithm is optimal for a k-fault-tolerant model. Our simulation results demonstrate that the proposed schedulers can efficiently improve energy savings when compared with previous works.
Energy-Efficient Fault-Tolerant Scheduling Algorithm for Real-Time Tasks in Cloud-Based 5G Networks
