Integrating Preemption Thresholds with Limited Preemption Scheduling

Abstract

The benefits of the limited preemption scheduling model serve to minimize preemption overhead while enabling cooperative scheduling between real-time tasks. Preemption point placement (PPP) algorithms are employed to select a suitable subset of preemption locations for limited preemption scheduling that optimize task worst case execution time. Similarly, preemption threshold scheduling enhances schedulability in a fully-preemptive environment by taking advantage of execution time slack in a task set by adjusting preemption thresholds to permit tasks to completely execute non-preemptively where possible. The ability to execute non-preemptively offers reduced cache related preemption delay (CRPD) further enhancing limited preemption schedulability. In this work, we integrate limited preemption scheduling using preemption placement with preemption threshold scheduling to realize further task set schedulability benefits. A case study using synthetically generated tasksets will demonstrate the significantly improved (up to a 30% increase in breakdown utilization) schedulability benefits of our proposed integrated PPP and optimal threshold assignment (OTA) algorithm.