Setting Priority Points for Global EDF-like Scheduling of Sporadic Tasks with Arbitrary Deadlines on Multiprocessors
MetadataShow full metadata
Under the well-studied global earliest-deadline-first (G-EDF) scheduling, the deadline of each job serves for two functionalities simultaneously. First, it specifies the response-time requirement of the job; second, it determines the priority of the job. The concept of the family of G-EDF-like (GEL) schedulers extends G-EDF by separating these two functionalities: the deadline of a job only serves as the first one, and another parameter, called the priority point of a job, serves as the second one.
This thesis studies the problem of GEL scheduling arbitrary-deadline sporadic tasks on multiprocessor platforms where every processor is identical. In particular, the focus is on the priority point setting and schedulability analysis under both preemptive and non-preemptive global earliest-priority-point-first (G-EPPF) scheduling. This work shows that the response times under G-EPPF scheduling can be upper bounded by the relative priority points as well as specification parameters of the tasks and the platform. Thus, the problem of setting proper priority points can be formalized as a linear program (LP), which can be solved efficiently by readily available LP solvers. The schedulability experiments demonstrate the effectiveness of the proposed method and its merits over the state-of-art for the schedulability analysis of arbitrary-deadline tasks.