作者: Anne Benoit , Aurélien Cavelan , Yves Robert , Hongyang Sun
DOI: 10.1007/978-3-319-17248-4_11
关键词: Job shop scheduling 、 Workflow 、 Energy consumption 、 Statistics 、 Computer science 、 Rollback recovery 、 General purpose 、 Distributed computing 、 Graph (abstract data type)
摘要: In this paper, we combine the traditional checkpointing and rollback recovery strategies with verification mechanisms to address both fail-stop silent errors. The objective is minimize either makespan or energy consumption. While DVFS a popular approach for reducing consumption, using lower speeds/voltages can increase number of errors, thereby complicating problem. We consider an application workflow whose dependence graph chain tasks, study three execution scenarios: (i) single speed used during whole execution; (ii) second, possibly higher any potential re-execution; (iii) different pairs speeds be throughout execution. For each scenario, determine optimal locations (and third scenario) objective. scenarios are then assessed compared through extensive set experiments.