摘要
目前基于软件容错模型的调度算法大部分都是针对如何提高主版本可执行性的预测精度,很少考虑如何提高主版本的完成率.基于这一问题,本文提出一种基于软件容错模型的实时调度算法(kernel),该算法有选择地重新执行某些失败的主版本并允许其运行在更高的优先级别上,尽最大努力地提高了主版本的完成率,改善了输出结果的计算精度.仿真实验表明,与同类算法相比,kernel算法在提高系统容错能力方面更为有效.
At present,almost all fault-tolerant scheduling algorithms are designed to improve the prediction of executable primaries,while less of them take how to enhance the success rate of primaries into account.Based on this problem,we present a new software fault-tolerant real-time scheduling algorithm,called kernel algorithm.This algorithm can choose some of the failed primaries to re-execute and allow them to be executed at some appropriate priority levels.The simulation experiments show that the kernel algorithm has much better fault resilience than the well-known algorithms so far.
出处
《小型微型计算机系统》
CSCD
北大核心
2010年第9期1804-1808,共5页
Journal of Chinese Computer Systems
基金
国家科技重大专项(2009ZX04009-013)资助
关键词
实时操作系统
调度算法
软件容错
主版本
副版本
real-time operating system
scheduling algorithms
software fault-tolerant
primary version
alternate version
