K-Priority Scheduling of Hard Real-Time Implicit-Deadline Periodic Task Systems on Uniprocessor

Abstract

The Rate Monotonic (RM) scheduling algorithm (static priority scheme) has an advantage in that it is simple to implement and incurs less run time overhead. It however has a disadvantage that it cannot guarantee 100% utilization of the processor for many tasks set having 100% tasks utilization. The earliest deadline first (EDF) scheduling algorithm (dynamic priority scheme) has a significant advantage in terms of processor utilization but increase in runtime overhead undermines this advantage. In this thesis, a dual priority algorithm is proposed. Each task has two static priorities – lower and upper band priorities with each priority band following the rate monotonic priority ordering. Initially each task executes with the lower band priority and promoted to the upper band priority after a particular (promotion point) time offset from the release of the task. We show by using simulation that 100% processor utilization is possible and our conjecture is that 100% processor utilization is possible.