【摘要】 在硬实时系统中,由于任务超时完成将会导致灾难性后果,因而硬实时系统具有严格的时间及可靠性限制条件.目前实时容错调度算法大部分针对硬件的容错,很少考虑软件运行的故障.提出了两种类似EDF(earliest deadline first)的软件容错的动态实时调度算法:PKSA(probing K-step algorithm)和CUBA(changing utilization-based algorithm).两种算法在任务执行过程中,通过若干试探性检测步骤,提高了任务可执行性的预测,尽可能地避免了任务早期的失败对后续任务的影响,因此提高了任务的完成率,同时也有效地减少了浪费的CPU时间片数量.通过实验测试,与目前所知的同类算法相比,具有更佳的调度性能——调度成本平衡性.更多还原

【Abstract】 A hard real-time system is usually subject to stringent reliability and timing constraints due to the fact that failure to produce correct results in a timely manner may lead to a disaster. Almost all fault-tolerant scheduling algorithms so far are designed to deal with hardware faults, less of those take possible software faults into account. This paper presents two software fault-tolerant real-time scheduling algorithms that are similar to EDF (earliest deadline first), called PKSA (probing K-step algorithm) and CUBA (changing utilization-based algorithm). The most important contribution of the algorithms is the probing of certain steps during the execution of tasks in order to prevent early failures in execution from triggering failures in the subsequent job executions. Therefore, the algorithms increase the successful percentage of task抯 completion, and meanwhile decrease the wasted CPU time slots. The simulation experiments show that the algorithms have much better trade-offs between scheduling costs and performance than the well-known algorithms so far.更多还原