Dealing with execution-overruns to improve the temporal robustness of real-time systems scheduled FP and EDF

Abstract

This paper presents a mechanism for execution-overrun management in preemptive real-time systems composed of sporadic tasks. We consider faults due to Worst Case Execution Time violations. Based on fault prevention, we determine the allowance of a task defined as the maximum extra duration that can be granted to a faulty task without compromising the timeliness constraints of the tasks. We propose a solution to implement the allowance principle that we call the Latest Execution Time (LET), defined for a task as the latest absolute time a faulty task can proceed with its execution without compromising the real-time constraints of all the tasks in the system. This mechanism enables to cope with execution overruns before a deadline miss and is not based on execution overruns handlers. We show how to determine the LET based on existing real-time feasibility conditions for Fixed Priority (FP) and Earliest Deadline First (EDF) schedulings. We compare the performances of the LET mechanism with FP and EDF scheduling to classical execution overrun management solutions.