Computing optimal communication schedules for time-triggered networks using an SMT solver

Multi-cluster systems with time-triggered networks are suitable for large safety-critical systems, which benefit from the inherent fault isolation and temporal predictability of the time-triggered paradigm. These networks depend on communication schedules that determine the global points in time for the transmission of messages with conflict-free paths through the switches, while satisfying real-time requirements and precedence constraints. On the basis of a state-of-the-art SMT solver, this paper introduces a novel optimal scheduler for time-triggered networks that is optimized for Boolean conditions and clause learning as required for efficient SMT solving. The ensuing improvements with respect to runtime, memory requirements and scalability are demonstrated by an experimental evaluation in the paper. Furthermore, we present techniques to parallelize the scheduling problem, which make the scheduler more efficient in distributed systems. Due to the lower runtime and memory requirements, the presented scheduler can even be suitable for dynamic computation of schedules in the embedded system itself as required for fault recovery by reconfiguration.