Boost-At-The-Tail: Work-Triggered Frequency Boosting for Fixed-Priority Scheduling (Artifact)

Abstract

Fixed-priority (FP) scheduling is widely used in safety-critical real-time systems due to its simplicity and analyzability, yet it may fail to schedule task sets whose worst-case response-time bounds exceed deadlines by small margins. Meanwhile, modern processors increasingly support short-term frequency boosting, providing additional processing capacity subject to thermal and power constraints. This paper introduces Boosted-FP, a fixed-priority scheduling framework that augments a given FP policy with controlled, limited frequency boosting. The key idea is to activate boosting based on the worst-case remaining execution demand of the currently executing job, thereby confining high-frequency execution to the tail of jobs. Per-task boost parameters are computed offline using standard fixed-priority response-time analysis under the chosen priority assignment, while boost activation decisions are made online using worst-case remaining-work information. We show that Boosted-FP preserves hard real-time guarantees by relating its execution behavior to that of a task set with reduced execution times. Leveraging the sustainability of fixed-priority schedulability analysis with respect to execution-time reductions, we establish that any task set schedulable under the modified execution bounds is also schedulable under the proposed framework. We evaluate the framework under a global boost-budget constraint and show experimentally that boost usage remains bounded and often well below the offline provisioned budget. Together, these results demonstrate that controlled, work-triggered boosting can safely extend the schedulable region of fixed-priority scheduling without abandoning static priorities.