Over recent decades, many studies have considered the development, assessment and refinement of scheduling algorithms for use in real-time embedded applications. Various studies have also considered the impact of variations in the interval between the executions of periodic tasks (i.e. jitter) on the behaviour of such systems. Despite interest in both of these areas, there has been comparatively little attention paid to the impact of scheduler implementation techniques on jitter behaviour. This is unfortunate because – as we demonstrate in the course of this paper – there is a ‘one-to-many’ mapping between scheduler algorithms and scheduler implementations, and even comparatively small changes in the scheduler implementation can have a significant impact on jitter behaviour. Throughout this paper, our focus is on implementations of a form of “cyclic executive” which is one of the simplest scheduling algorithms in widespread use. The results presented demonstrate that – even for this very simple scheduling algorithm – implementation decisions can have a significant impact on both jitter behaviour and on resource requirements. We would expect that the results obtained would also apply to more complicated algorithms: indeed, as the algorithms grow more complicated, we would expect that the number of implementation options would increase, with a corresponding increase in the jitter variation.

real-time, scheduling algorithm, scheduler implementation, jitter, time-triggered co-operative, cyclic executive, on-line schedule, off-line schedule