18th IEEE Symposium on Real-Time Distributed Computing (ISORC),
2015 · doi: 10.1109/ISORC.2015.10
Fault-tolerant distributed embedded systems have to react properly on the occurrence of faults in order to avoid harm to the system or its environment. Faulty system resources have to be isolated from the remaining system. Hence, these resources become unavailable, leading to a decreasing number of available resources and input data. In such cases, mechanisms like graceful degradation may be applied to ensure that the system does not turn off completely, but degrades its provided set of functional features gracefully. It must be ensured that the remaining intact resources are efficiently used to execute at least those features, which are required to behave fail-operational. In this paper, we investigate deployments of mixed-critical software components to a fault-tolerant system platform. We introduce a formal model of software components and their publish/subscribe based communication channels. We use this model to analyze the graceful degradation of the system in different scenarios of failing execution hardware. This includes also the explicit deactivation of software components due to unavailable required input data. Our analysis is based on using an SMT solver and contributes to guarantee that all requirements with respect to fail-operationality are met by the system design. The approach is evaluated by an example and a scalability analysis.