An Optimal Discrete-Time Based Boundary Fair Scheduler for Sporadic Tasks in Multiprocessor Real-Time Systems∗

Several optimal scheduling algorithms have been studied recently for various real-time tasks running on multiprocessor systems with continuous as well as discrete time. However, the existing optimal schedulers for sporadic tasks may incur very high scheduling overhead due to either excessive number of scheduling points (e.g., for the Pfair-like scheduler) or arbitrarily small allocation for tasks (e.g., for the continuous-time based DP-Fair scheduler). In this paper, extending the boundary fair (Bfair) scheduler that is optimal for periodic tasks, we investigate an optimal scheduling algorithm for such sporadic tasks with implicit deadlines in multiprocessor systems. First, considering the irregular arrivals of sporadic tasks, we discuss how to determine the next boundary time (i.e., the deadline of a task) to minimize the scheduling points. Then, an example is presented to illustrate the major scheduling steps that incorporate the dynamic arrivals of sporadic tasks, followed by the formal presentation of the scheduling algorithm. We point out our future work in the end.