Adaptive Checkpointing Schemes for Fault Tolerance in Real-Time Systems with Task Duplication

Dynamic adaptation techniques based on checkpointing is studied in this paper. Placing store-checkpoints and compare-checkpoints between CSCP (store-and-compare-checkpoint), we first present adaptive checkpointing schemes in which the checkpointing interval for a task is dynamically adjusted on line. Introducing the overheads of comparison and storage, the average execution times to complete a task for proposed schemes are obtained, using renewal equations. Further, we have dis- cussed analytically the optimal numbers of checkpoints that minimize the average execution times. We then extend proposed schemes to a set of multiple tasks in real-time systems. Simulation results show that compared to previous method, the proposed approach significantly increases the likelihood of timely task completion.