摘要
针对多处理器硬实时系统运行过程中任务出错问题,提出一种基于主副版本策略的固定优先级全局容错调度算法FTGS-BD(fault tolerant global scheduling with backup delay)。该算法使用主动副版本和被动副版本,在保证实时性的前提下根据任务需求和硬件性能尽可能的推迟主动副版本的运行,并在不需要副版本响应时回收分配给副版本的资源,从而减少实现容错所需的代价。仿真结果表明,和仅使用被动副版本的全局容错调度算法相比,在调度相同的任务集时,FTGS-BD最多可以减少20%的处理器资源需求,平均减少12%。FTGS-BD能够应用于主副版本总使用率最大值大于1的任务集。
In hard real-time system, faults happen every now and then. With respect to the criticality of this kind of system, the capability of fault tolerance is necessary. A fault tolerant scheduling algorithm is capable of guarantee- ing timing requirements of tasks even when faults occur. A primary-backup based fault tolerant global scheduling algorithm FTGS-BD （fault tolerant global scheduling with backup delay） is proposed in this paper. FTGS-BD pu- shes the execution of active backup as late as possible and reclaims the resource distributed to backups when it is no longer needed. By this approach the additional cost for achieving fault tolerance is significantly reduced. The simu- lation result shows that, comparing with the scheduling algorithm which only implements passive backups, FTGS- vBD reduces processors requirement 20% max and 12% on average. FTGS-BD could also be adopted when high utilization tasks exist.
出处
《电子测量与仪器学报》
CSCD
北大核心
2015年第9期1302-1309,共8页
Journal of Electronic Measurement and Instrumentation
基金
国家自然科学基金(61370088)
国家国际科技合作专项(2014DFB10060)资助项目
关键词
多处理器
硬实时系统
主副版本
容错调度
全局调度
muhiprocessor
hard real-time systems
primary-backup
fault tolerant scheduling
global scheduling
