First-Fit Semi-partitioned Scheduling Based on Rate Monotonic Algorithm

Abstract

Semi-partitioned scheduling is a new approach for allocating real-time tasks to processors such that utilization is enhanced. Each semi-partitioned approach has two phases, partitioning and scheduling. In partitioning phase, tasks are assigned to the processors. In this phase, some tasks are probably split into several subtasks and each assigned to a different processor. The second phase is the policy to determine how to schedule tasks on each processor. The main challenge of semi-partitioned scheduling algorithms is how to partition and split tasks by which they are safely scheduled under the identified scheduling policy. This paper proposes a new semi-partitioned scheduling algorithm called SRM-FF for real-time periodic tasks over multiprocessor platforms. The scheduling policy used within each processor is based on rate monotonic algorithm. The partitioning phase of our proposed approach includes two sub-phases. Task splitting is done only in the second sub-phase. In the first sub-phase, processors are selected by a first-fit method. The use of first-fit method makes SRM-FF create lower number of subtasks in comparison to previous work hence the number of context switches of subtasks and overhead due to task splitting are reduced. The feasibility of tasks and subtasks which are partitioned by SRM-FF is formally proved.