Adaptation for Energy Saving in Time-Triggered Systems Using Meta-scheduling with Sample Points.

Time-triggered systems offer significant advantages in embedded applications due to temporal predictability, implicit synchronisation and avoidance of resource contention. However, runtime adaptation of system services in these systems is motivated by energy efficiency without detriment to system performance. Too frequent adaptation result in more communication overhead. This work introduces a meta-scheduling technique with sample points to compute adapted static schedules for energy saving in time-triggered systems. An offline meta-scheduler optimises static schedules by applying slack events for energy saving at global periodic points in a schedule. The meta-scheduler maps the adaptation points to the runtime sampling period of adaptation. Slack events are reported synchronously by adaptation units at runtime, and adaptation is achieved through the aligned switching of component schedules facilitated by a Fault-Tolerant Agreement Protocol (FTAP). The meta-scheduler computes an MSG which holds all adapted schedules and describes the runtime switching of schedules based on reported slack events. The increased overhead due to periodic adaptation of the system schedule and energy saving of the meta-scheduling technique are evaluated. Results show a reduction in energy consumption compared with a base schedule while highlighting the trade-off between increased communication overhead and energy-saving.