Probabilistic Analysis of Real-time Scheduling of Systems Tolerating Multiple Transient Faults

The influence of computer systems in human life is increasing and thereby increases the need for having reliable, robust and real-time services of computer systems. Avoidance of any catastrophic consequences due to faults in such systems is the main objective. This paper addresses the problem of finding a probabilistic measure of schedulability of real-time systems tasks in the presence of multiple transient faults. The main approach used in this paper is employing Temporal Error Masking (TEM) technique to achieve Node Level Fault Tolerance (NLFT) within the least common multiple of periods of a set of pre-emptively scheduled periodic tasks with at most f transient faults. Rate Monotonic (RM) scheduling is used to observe how the probability of system success, that is, the probability of meeting deadlines for all tasks is affected with the worst-case distribution of f faults