Antinomy between schedulability and quality of control using a feedback scheduler

Abstract

This paper addresses system control quality in a real-time context. Comparing the impact of sampling period reduction by a Feedback Scheduler, using three different processor utilization bounds, we stress the antinomy of control quality and schedulability. We then measure statistically the impact of scheduling artefacts (task jitters, input/output delays) to outline the main factors causing this antinomy.

A PID controller based on the processor utilization artifact within a real-time multitask control system

“…Consequently, the present study confirms that the QC highly depends on the sampling and the input–output latencies. This finding confirms the antinomy statistically proved in Zakaria et al 13…”

“…The dynamic of each servomotor is described by the continuous time transfer function G ( s ) = ( 1000 ) / ( s ( s + 1 ) ) . For an emulation with n jobs of durations C i ( k ) k ∈ 1 . . n , such as in Zakaria et al, 13,14 a Weibull random generator is executed during a simulation time t sim = 5 s with the characteristics (in ms) shown in Table 2 for the control tasks τ i ∈ 1 . . 3 . The generator is configured by three parameters: a localization l , a shape factor μ , and a scale factor λ to provide the worst execution time of 4 ms and the best execution time of 3.3 ms. We set the FBS period to 5.5 ms to readjust tasks periods as soon as the estimated execution time changes; in such a way as to get the utilization rate equal to the L & L bound 26 where τ i +…”

“…On the other hand, the authors in Zakaria et al 13 found that it is not always interesting when the FBS based on the processor utilization bound is used because of the correlation between the sampling period and the latency, which may be interpreted as an antinomy between scheduling constraints and control design rules. All previous solutions are based on algorithms that may or may not be decidable, where their complexity conceivably becomes relevant.…”

“…In exceptional situations, a performance cost function can be associated with a soft task (control degrades with response time), where the execution time can be communicated by feed-forwarding the mode-change information to the scheduler. 9 On the other hand, the authors in Zakaria et al 13 found that it is not always interesting when the FBS based on the processor utilization bound is used because of the correlation between the sampling period and the latency, which may be interpreted as an antinomy between scheduling constraints and control design rules. All previous solutions are based on algorithms that may or may not be decidable, where their complexity conceivably becomes relevant.…”

Performance Optimization of Control Applications on Fog Computing Platforms Using Scheduling and Isolation

“…The scheduler gets delay and jitter feedback to change the period of the tasks concerning the QoC and workload management. The same approach is used in [56], to schedule tasks concerning the QoC with feedback from scheduling. In this work, the period of tasks is changed regarding the feedback.…”