Checkpointing Workflows à la Young/Daly Is Not Good Enough

This article revisits checkpointing strategies when workflows composed of multiple tasks execute on a parallel platform. The objective is to minimize the expectation total execution time. For single task, Young/Daly formula provides optimal period. However, many simultaneously, risk that one them severely delayed increases with number tasks. To mitigate this risk, possibility checkpoint each task more often than strategy. But it worth slowing down extra checkpoints? Does make difference globally? answers these questions. On theoretical side, we prove several negative results for keeping period concurrently, and design novel guarantee an efficient high probability. practical report comprehensive experiments demonstrate need go beyond wide range application/platform settings.