Performability guarantee for periodic tasks in real-time systems

Performability is an important parameter in safety-critical real-time systems. This parameter is defined as the joint consideration of two other important parameters, i.e., reliability and performance. This paper proposes a schedulability condition that guarantees a desired level of performability under various working conditions for real-time systems. The basic idea underlining this condition is to select a subset of schedulable tasks and manage their slack times to satisfy a desired performability level. The proposed condition is evaluated on a hard real-time system that employs the Rate-Monotonic (RM) scheduling algorithm and uses the re-execution mechanism to improve the reliability. Evaluation results reveal that by employing the condition, the level of performability of the system is always greater than the desired performability. In addition, it yields, on average, 1% improvement in system performability in comparison with traditional schedulability conditions, while the actual failure rate is greater than the expected rate. This amount of performability improvement is significant for safety-critical real-time systems. (C) 2014 Sharif University of Technology. All rights reserved.