一种空间多核操作系统容错调度算法

A fault-tolerant scheduling algorithm for space multi-core operating systems

目前计算机系统逐步采用多核处理器来提升性能,空间操作系统如何管理多核资源是发挥处理器性能的关键。在航天等安全关键领域中,采用固定点任务与定期任务混合调度,在保证可靠性的前提下提高效率。现有针对混合任务模型的多核调度算法仅考虑任务分配问题,没有考虑到系统中某一核心出现故障时如何进行容错。FT-RTA算法是一种空间多核操作系统容错调度算法,当一个核心上出现瞬时故障,将故障核心上在故障时间段内的所有任务迁移至正常核心上执行,使计算机系统不会感知到此次核心故障,成功屏蔽故障。经过实际应用中的典型参数验证,算法可以成功屏蔽核心故障,进行系统无感知的容错。

Currently,computer systems are gradually adopting multi-core processors to improve performance,and how the space operating system manages multi-core resources is the key to leveraging processor performance.In safety-critical areas such as aerospace,hybrid scheduling of fixed-point tasks and periodic tasks is used to improve efficiency while ensuring reliability.Existing multi-core scheduling algorithms for the hybrid task model only consider the task allocation problem,and do not take into account how to be fault-tolerant when one of the cores in the system fails.The FT-RTA algorithm is a fault-tolerant scheduling algorithm for space multi-core operating systems.When there is a transient failure on one core,the algorithm migrates all the tasks on the faulty core in the faulty time period to be executed on normal cores,so that the computer system does not perceive this core failure,successfully shielding the fault.The algorithm can be successfully shielded from core failures and carry out system-unaware fault tolerance after typical parameter verification in real applications.