What is the meaning of preemption in utility-based real-time scheduling?

I. PRELIMINARIES Time utility function (TUF) scheduling goes beyond the starttime-deadline notion to express tasks’ temporal constraints. The goal of the scheduler is to maximize system utility under the assumption that tasks aggregate a given amount of utility to the system as a function of when they execute. Work available in the literature include the characterization of time utility functions to express applications’ requirements [5], [7], and several proposals of scheduling algorithms for increased utility accrual [8], [9], [3] which differ in their supported utility functions, overhead, etc. Most work on TUF scheduling, including all previously mentioned scheduling algorithms, do not consider preemption. Many simple embedded system do not support preemption, and some applications, like network packet scheduling, are inherently non-preemptive. However, if allowed, preemption opens up possibility for more efficient resource utilization, and being able to explore it is advantageous. We believe that the scarcity of work considering preemption reflects the lack of understanding of the impact of preemption on the utility accrual. Moreover, preemptive scheduling imposes extra challenges such as context switch overhead, which may lead to poor and inefficient resource utilization.