Probabilistic Evaluation of Checkpoint-Based Fault Tolerance in Real-Time Systems

This paper discusses the execution of a real-time program in an unreliable environment, which can ultimately lead to a failure, and then it is repaired, but all the work done is lost when the system crashes and the calculations should start again. Practitioners are now interested in such fault tolerance technique as checkpointing, which should introduce small overhead and is aimed at reducing the amount of work that is lost upon failure of the system. In connection with this interest, stochastic model was presented and such random variable as bounded accumulated unproductive time was considered. By using the Laplace-Stieltjes transform the functions for the probability that real-time program'll meet deadline (with and without checkpoints) were obtained.