Energy-aware Scheduling Algorithm With Fault-Tolerance In Real-Time Systems

The energy consumption of real-time systems has been the research focus. There are software and hardware methods for energy-saving. So far, most of software techniques are based on DVS (Dynamic Voltage scaled). However, recent research shows that DVS has a direct and adverse effect on the transient fault rates, and will reduces the reliability of the system. Therefore, energy conservation through the DVS while maintaining system reliability has become a new research branch.By introducing fault-tolerant technology, system reliability can be ensured. Fault tolerance is mainly achieved by redundant design. Usually during energy saving by software techniques, slack time is the main redundancy resource. However, Slack time is a sensitive concept in real-time system, especially when using DVS for energy-saving, which becomes a key factor to balance energy consumption and reliability of the system.Based on the understanding of a variety of energy conservation and fault-tolerant related technologies, focus on the RA-PM (Reliability-Aware Power Management) algorithm, which considers these two technologies together, this study expands and improves it. For the dynamic RA-PM algorithm, at first we expand the original algorithm. Considering when there is only one task in the ready queue, we propose the concept of wrapper tasks"reorganization"to avoid slack waste. Then we reconsider the timing of the wrapper task's deadline extension in order to increase the utilization of idle, by which we improve RA-DPM algorithm. For static RA-PM algorithm, we introduce the concept of dummy task and change static slack to dynamic slack, and then we can manage static slack with wrapper task like dynamic slack and make full use of static idle slots.Experimental results show that compared with the original algorithm our algorithm can achieve better energy efficiency and keep the same reliability as the origin.