Preemptive online scheduling in the computing continuum

Due to the high heterogeneity of the resources across the Computing Continuum, it becomes increasingly difficult to efficiently schedule the modern IoT applications, while guaranteeing the lowest possible execution time. Traditional scheduling approaches consolidate the IoT applications on a restricted number of Computing Continuum resources and are conducted offline, leading to resource over-provisioning and performance bottlenecks. This paper introduces an online preemptive algorithm for scheduling IoT applications, named PASIoT, tailored explicitly for the Computing Continuum. To validate the behavior of PASIoT, we compared it with a state-of-the-art scheduler over a real-life testbed. The evaluation results showed that the PASIoT approach achieved 10 - 50 % lower IoT application completion time than related works.