分布式仿真任务的最坏反应时间分析

The Worst-Case Response Time Analysis for Distributed Simulation Task

最坏反应时间是进行任务可调度性分析的标准之一。我们对采用被动备份的PB(Primary-Backup)模型进行容错的仿真任务的最坏反应时间分别进行了分析,其中主、副版本仿真任务之间的同步分别采用重新执行和共享检查点文件,给出了相应最坏反应时间的计算公式。仿真试验结果表明基于共享检查点文件进行主、副版本任务同步会获得更短的最坏反应时间。

Worst-case response time is one of the most important schedulability analysis criteria. In simulation system, the fault tolerance is based on passive duplication PB model, and there are two methods, re-execution and checkpoint file sharing, which can be used to synchronize the primary and backup version simulation task. The formulas were put forward for computing the worst case response time of task when external or internal error occurs, in which the fault recovery of task is based on re-execution or checkpoint file sharing methods. The final simulation test shows that the checkpoint-based fault recovery makes the simulation task have shorter worst case response time and make the task set more schedulable.