A Hybrid Multiprocessor Scheduling Approach for Weakly Hard Real-Time Tasks

There are two major strategies to schedule real-time tasks in multiprocessor systems; partitioning and global scheduling. The partitioning approach has acceptable overhead but cannot guarantee to be optimal. The global approach can provide this guarantee but it has considerable overhead. Thus, a multiprocessor real-time scheduling approach for weakly hard real-time tasks is proposed that employs hybrid scheduling. Studies have shown that current multiprocessor scheduling of weakly hard real-time tasks used imprecise computation model based on iterative algorithms. This algorithm decomposed into two parts; mandatory and optional, unfortunately, the result analysis is precise only if its mandatory and optional parts are both executed. Even, the use of hierarchical scheduling algorithm, such as two-level scheduling under PFair algorithm may cause high overhead due to frequent preemptions and migrations. In this paper, an alternative scheduling approach will be proposed, which is, its combines elements of the two well-known multiprocessor scheduling approaches. It aims to employs benefits and advantages of the partitioning and global scheduling. Accordingly, the proposed hybrid multiprocessor real-time scheduling is use the best algorithm of each of partitioning and global approaches, R-BOUND-MP-NFRNS and RM-US (m/3m-2) with multiprocessor response time test. Schedulability experiments and simulation results using Matlab show the proposed hybrid multiprocessor scheduling approach to be effective for weakly hard real-time tasks.