Task Level Multimode Real-time Task Scheduling Research

Real-time systems are referred to as the computing system which can provide therequired level service in a specified time. The task in these systems has not onlyexecution time requirements, but also the quality requirements. In real-time system,the task model and the corresponding scheduling policy can make the task completesmoothly. It is key to establishing an effective task model and des igning a flexiblescheduling strategy in real-time system.Multiframe task is a kind task model in which one task has various executiontimes, but the application type of this kind of task model is limited. Our research ofthis paper focuses on scheduling problem of tasks with various execution timecharacteristics. A multimode task model is proposed in which every task has varietyof versions. Based on the research of the multiframe task scheduling and the dynamicreal-time scheduling, this paper intends t o apply the multimode task into dynamicreal-time scheduling. The specific research works are as follows:At first, for the scheduling of the periodic multimode task, the task executionmodel selection has the direct effect on the task-set schedulability and systemperformance. Considering the actual execution time of the task, we apply themultimode task into dynamic real-time scheduling, and raise a two phases slackreclamation scheduling algorithm in this paper. This scheduling algorithm iscomposed of offline phase and online phase. In offline phase, the instance executionmode selection and the scheduling algorithm selection is determined, and in onlinephase, the slack can be utilized efficiently by the characteristics of this task mode.Under the premise that all tasks meet deadlines, this scheduling algorithm canefficiently use system resources and improve system performance.In addition, for the hybrid multimode task system composed of periodic andsporadic multimode tasks, this paper considers both the system performance and theoverload caused by sporadic multimode task. On the foundation of the period icmultimode task scheduling, a hybrid task model overload processing mechanism wasproposed, which includes sporadic task acceptance test and overlo ad processingmechanism based on task execution model selection. For the task execution modeselection mechanism, an improved method based on imprecise computation modeloverload processing was proposed. In order to estimate the effectiveness of the algori thms, this paper presentssimulation test for the task scheduling strategy of the periodic multimode task andhybrid multimode task respectively. The results reveal that multimode task using EDLscheduling algorithm has obvious advantages, and slack reclamation mechanism withselectivity characteristic under corresponding conditions can provide higher accuredreward value. Using the improved overload processing mechanism can get a better hitvalue ratio.