When it comes to real-time systems, the schedulability of a given task set will always be more likely if preemption of the running task is permitted. Preemption does, however, have some drawbacks, one of which is time spent switching tasks, while another, according to the authors, is the over-estimation of the worst-case execution time for tasks due to the variability that preemption introduces. Limited preemption is a midway point where a task may only be preempted at certain points. One obvious appeal is the simplified design permitted for critical section synchronization.

The authors look at the impact on scheduling feasibility that the limited preemptive paradigm provides. Much of the paper is dedicated to deriving the simple (and intuitive) limit that the longest non-preemptive section over all tasks must be less than half of the shortest deadline once the central processing unit (CPU) speed-up has been taken into account. Admittedly, the authors phrase it in terms of the required CPU speed-up to ensure schedule feasibility given non-linear scaling of the execution time with processor speed. They do, however, eventually require the scaling to be linear to be able to provide an upper bound.

The final part of the paper looks at the impact that multiple cores or processors have on the schedulability of a given task set. Once again, the final derived lemma can be simply stated that as long as the shortest deadline is less than or equal to the longest non-preemptive execution segment, no more than three processors are required to ensure schedulability.