First implementation and test of reintegration mechanisms for node replicas in the FT4FTT Architecture

摘要

Distributed Embedded Control Systems (DECSs) used for critical applications must usually abide by strict real-time and dependability requirements. Correspondingly, the FT4FTT project proposes a complete fault-tolerant (FT) architecture for RT DECSs. The Flexible Time-Triggered Ethernet (FTT-Ethernet) communication protocol fulfills the RT requirements, while the FT mechanisms added on top of it, which are based on channel duplication and active replication of nodes, provide the FT behaviour. Temporary faults affecting the channel or the nodes, which are the most probable type of faults in DESs, can manifest in such a way that a node replica loses its coordination with the others and, thereby, it also loses its communication and/or computation capability from then on, leading to attrition of the redundancy initially provided by the active replication of nodes. This paper describes the implementation and test of specific mechanisms that are devised to determine which replicas are temporarily faulty and to promptly reintegrate them.