Research On Multi-core Real-Time Task Scheduling Based On Task-Split

In recent years, multicore platforms have proliferated in the market and research has shifted to the more complex and computing-intensive tasks. However, the complexity of multicore processors causes the situation that tasks become even more inconsistent, unpredictable, and varies. Due to dependencies exist among different tasks, task scheduling becomes uncoordinated, which causes them waiting for each other and missing the deadline. Therefore, traditional real-time task scheduling based on independent tasks faces enormous challenges and the new scheduling algorithm taking dependent tasks into account is increasingly becoming new directions for research.Based on the analysis and study of existing real-time task scheduling algorithm on multi-core platforms, a partitioned fixed-priority real-time scheduling with dependent tasks-split on homogeneous multi-core platform is proposed in this paper, namely TDTD/TS (Tree Dependent Task Dispatching/Task Splitting). This strategy deals with hybrid real-time tasks (dependent and non-dependent tasks) on multi-core platforms, which meets the requirement of real-time applications in strict time. Considering the influence in dynamics and dependency of tasks, we design the scheduling hierarchically into two phases:TDTD (Tree Dependent Task Dispatching) and TDTS (Tree Dependent Task Split). TDTD converts dependent tasks into a series of sequential jobs and obtains the interrelated subtasks path as well as synthetic deadlines through the B-tree task model. Then in TDTS, dependent task in idle-wait state is qualified to split and all blocked sub-tasks would preempt CPU immediately to guarantee deadlines. With regard to utilization bound analysis, the new algorithm is proved to offer superior performance guarantee69.31%.The simulations and experimental results on gem5and test machine show that the protocol has a good schedulability and QoS performance compared with existing algorithms, especially in time rate, SLR (Schedule Length Ratio) and Speedup. It proves that the algorithm TDTD/TS can provide a higher level of usability and efficiency, which has a comparative advantage.