Checkpointing strategies to protect parallel jobs from non-memoryless fail-stop errors
Anne Benoît, Lucas Perotin, Yves Robert, Frédéric Vivien · HAL (Le Centre pour la Communication Scientifique Directe) · 2022
This paper studies checkpointing strategies for parallel jobs subject to fail-stop errors. The optimal strategy is well known when failure inter-arrival times obey an Exponential law, but it is unknown for non-memoryless failure distributions. We explain why the latter fact is misunderstood in recent literature. We propose a general strategy that maximizes the expected efficiency until the next failure, and we show that this strategy is asymptotically optimal for very long jobs. Through extensive simulations, we show that the new strategy is always at least as good as the Young/Daly strategy for various failure distributions. For distributions with a high infant mortality (such as LogNormal 2.51 or Weibull 0.5), the execution time is divided by a factor 1.9 on average, and up to a factor 4.2 for recently deployed platforms.