Fixed priority scheduling with deadlines prior to completion

Abstract

Standard analysis for fixed priority scheduling assumes that all the computations for each task must be completed by the task's deadline. In practice this is rarely the case. The deadline is more correctly associated with the last observable event of the task, interval events and kernel overheads (such as the context switch away from the task) can occur after the deadline. New analysis is presented that caters for this situation. The result is enhanced schedulability without modification to the preemptive behaviour of the run-time system.<>