基于检查点间隔优化的容错实时系统可调度性

Schedulability of fault-tolerant real-time system based on checkpoint interval optimization

针对实时系统容错模型只能容忍任务执行时一次故障发生的缺陷,基于卷回恢复容错模型,提出了任务局部最优检查点间隔求解算法,使实时系统任务能够容忍任务执行过程中多次故障的发生;进而基于局部最优检查点间隔提出任务集检查点间隔全局优化(CIGO)策略。基于不同的检查点间隔优化机制,结合实时系统任务最坏响应时间的计算公式进行系统可调度性分析。仿真实验表明,与局部最优检查点间隔相比,基于全局优化检查点间隔的实时系统容错能力有显著提升。

To overcome the defect of the real-time system fault-tolerant model that could only tolerate one fault occurred during the task execution process, a task Local Optimum Checkpoint Interval （LOCI） iterative algorithm was proposed. The algorithm is based on the checkpoint and rollback recovery fault tolerant model to make the task of the real-time system could tolerate multiple faults occurred during the task execution process. Then the Checkpoint Interval Global Optimization （CIGO） algorithm of the task set was proposed. The sehedulability of the real-time system was analyzed based on the different checkpoint interval optimization mechanisms according to the calculation formula of the task worst case response time. The simulation results show that the strategy based on the GIGI can significantly improve the system fault resilience compared to that based on the LOCI.