Fault-tolerant hierarchical real-time scheduling with backup partitions on single processor

Abstract

The resource partitioning has been suggested to provide efficient composition of multi-threaded real-time applications. Partitioning can provide reliable and flexible software upgrade as partitions are strongly isolated in terms of resources. However, there are always possibility of experiencing software faults while operating on a real plant. To avoid entering a hazardous state due to a partition that is yet to be fully verified, we can deploy a backup partition that may implement inefficient algorithms or limited features but is verified with respect to reliability. The backup partition performs failover to carry out missions of the corresponding primary partition when a software fault is detected. There have been significant researches for fault-tolerant real-time scheduling but considerations for partitioned systems have not been studied. In this paper, we extend the resource model for hierarchical real-time scheduling to support primary and backup partitions. Our model can support context-dependent and context-independent tasks in the backup partition efficiently. In addition, we provide the schedulability analysis for suggested model.