Scheduling Dynamic Task-Sets in Time-Triggered Real-Time Systems

The time-triggered paradigm of activation has shown prominence in supporting safety critical applications deployed over distributed systems of networked nodes, particularly those with complex temporal constraints, e.g., distributed end-to-end deadlines.By triggering activities at predetermined points in time and letting them execute exclusively in dedicated time intervals (time-slots), extra-functional properties w.r.t.timeliness and dependability, e.g., determinism, reliability and temporal partitioning, can be efficiently guaranteed.In native time-triggered implementations, supporting dynamically changing task sets during runtime is challenging since all possible execution scenarios must be accommodated with a-priori assigned time-slots.This paper addresses the problem of efficiently handling task sets dynamically changing over time in time-triggered systems.In particular, we propose the Dynamic Task-Set Scheduler (DTSS) algorithm, which enables tasks to be admitted, removed, or controllably skipped without losing the extra-functional properties of the native time-triggered implementations.Hence, it safely permits flexible operation in the otherwise strict slot-table-driven timetriggered systems.DTSS is founded on an exact feasibility analysis, tailored specifically for dynamically changing task sets.By leveraging the minimum and sufficient number of constraints for the feasibility analysis, DTSS keeps the computational overhead low.Experimental results demonstrate overhead improvements compared to state-of-the-art methods for the same execution scenarios.