Optimal Checkpointing Strategies for Iterative Applications

This work provides an optimal checkpointing strategy to protect iterative applications from fail-stop errors. We consider a general framework, where the application repeats same execution pattern by executing consecutive iterations, and each iteration is composed of several tasks. These tasks have different lengths checkpoint costs. Assume that there are n task i , 0 ? i <; has time t cost c . A naive would after task. Another at end iteration. inspired Young/Daly formula for ?{2 ?c xmlns:xlink="http://www.w3.org/1999/xlink">ave } seconds, ? MTBF average time, current (and repeat). strategy, also formula, select xmlns:xlink="http://www.w3.org/1999/xlink">min with smallest every p th instance task, leading period T, T = ? xmlns:xlink="http://www.w3.org/1999/xlink">i=0 xmlns:xlink="http://www.w3.org/1999/xlink">n-1 per One choose so ? obey formula. All these strategies suboptimal. Our main contribution show globally periodic, design dynamic programming algorithm computes pattern. may well many tasks, this across iterations. through simulations, both synthetic real-life scenarios, outperforms strategies.