A fault-tolerant dynamic scheduler for distributed hard-real-time systems

A dynamic run-time scheduler is proposed that enhances the effects of a pre-run-time scheduling algorithm for real-time applications. The tasks are of a periodic nature and each task has two schedulable versions: primary and alternate. The primary produces an accurate result while the alternate produces an approximate result but takes less time and should be scheduled if the primary fails to meet the deadline. The objective of the dynamic scheduler is to maximize the number of primaries that are scheduled. The scheduling algorithm and performance results for different failure rates are also given. The performance results show that, for lower failure probability of the primaries scheduled during pre-run-time, the algorithm succeeds in scheduling a higher number of primaries during run-time