Abstract
The notion of preemption threshold is initiated in the industry to provide flexibility for real-time and embedded system designs. However, it also brings new contents to scheduling theory. Historically, the scheduling model are divided into two categories: preemptive and non-preemptive. In the context of fixed-priority scheduling, feasibility of a task set with non-preemptive scheduling does not imply the feasibility with preemptive scheduling and vice versa. We demonstrate a generalized fixed-priority scheduling model using the notion of preemption threshold, bridging the gap between preemptive scheduler and non-preemptive scheduler, such that both of them are included in the new model as extreme cases. With this model, the schedulability of a task set is improved compared with both preemptive and non-preemptive scheduler. In this paper, we develop equations for computing worst-case response time using the concept of level-i busy period. Some useful results about the generalized model are presented and an algorithm for optimal assignment of priority and preemption threshold is designed based on these results.