PC-Fair: Fair Scheduling of Precedence-Constrained Periodic Tasks in Multiprocessor Systems

The Pfair scheduling algorithms are known to be optimal for scheduling periodic real-time tasks on multiprocessor systems. However, one of the limitations of most prior work on the Pfair scheduling is that precedence constraints among tasks are not considered. We thus propose a PC-Fair (PrecedenceConstrained Fair) scheduling algorithm to overcome this limitation. The novelty of our algorithm comes from its two-level hybrid design. First, we adopt a Pfair scheduling algorithm to allocate a fair amount of CPU time to a group of tasks that have precedence constraints among them. Then, we revise an existing list scheduling algorithm and use it to schedule tasks inside each group while considering the precedence constraints. We also present an exact schedulability test for the proposed algorithm. Finally, we have implemented the algorithm and performed a series of simulations. The experimental results show strong evidence that the input-to-output response time of the schedule generated by our algorithm approaches to the optimal response time when the total utilization is high.