Seventh International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT'06)
Prediction-Table Based Fault-Tolerant Real-Time Scheduling Algorithm
Taipei, Taiwan
December 04-December 07
ISBN: 0-7695-2736-1
In order to predict accurately whether primary versions of real-time tasks is executable in software fault-tolerant module, a new algorithm, PTBA, Prediction-Table Based Algorithm, is presented. PTBA uses prediction-table to predict whether a host primary can meet its pre-deadline. Prediction-table contains the pre-assignment information of tasks between the current time and the alternates' notification time. If the prediction result shows that host primary has not enough time to execute, it will be aborted. Otherwise, prediction-table is referenced to schedule tasks with low overhead. The novelty of PTBA is that it schedules primaries according to their corresponding alternates' notification time and has no extra scheduling overhead in prediction-table mode. Simulation results show that PTBA allows more execution time for primaries and wastes less processor time than the well-known similar algorithms. PTBA is appropriate to the situation where the periods of tasks are short and software fault probability is low.
Citation:
Dong Liu, Chunyuan Zhang, Rui Li, "Prediction-Table Based Fault-Tolerant Real-Time Scheduling Algorithm," pdcat, pp.144-149, Seventh International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT'06), 2006