Design And Implementation Of Mission-critical Scheduling Platform Based On Heterogeneous Multi-core Processor

With the increasingly wide application of intelligent wearable devices in life,military and medical fields,how to improve the reliability of embedded system to better apply to more application scenarios has become a problem worthy of study.Firstly,this paper analyzes the application scenarios of wearable devices,analyzes the problems of wearable devices in practical application scenarios in detail by combining with some existing fault-tolerant scheduling algorithms,and proposes a task-oriented fault-tolerant scheduling platform to provide fault-tolerant protection for key tasks at the software level.Combining the requirements of wearable devices for size,power consumption and operating efficiency,a hardware platform combining RK3399 chip and Zynq ZC706 was built,and a task processing module taking edge detection algorithm and image enhancement algorithm as examples was realized.A fault tolerant scheduling module based on dual mode redundancy is implemented to ensure the reliability of critical tasks.Combined with the characteristics of heterogeneous multi-core architecture,a task synchronization mechanism between processors is proposed,through which the fault-tolerant scheduling algorithm is improved,and the running time of the fault-tolerant algorithm is shortened without affecting the fault-tolerant performance.At the same time,the task model based on DAG is analyzed from the perspective of task reliability.The transient faults of these sub-tasks in the running time will have a chain reaction,which will adversely affect the reliability of the system.To solve this problem,the concept of clustering based fault tolerant scheduling algorithm and balance coefficient is proposed.By combining the clustering algorithm with the table scheduling algorithm,different fault tolerant strategies are provided for different tasks in the DAG graph to improve the performance of the scheduling algorithm.Finally,the comparison experiment of fault tolerant scheduling algorithm before and after the improvement is carried out.The experimental results show that the scheduling module with stable and effective fault tolerance performance is implemented in the heterogeneous multicore architecture based on RK3399+Zynq,which takes edge detection and image enhancement as computing tasks.The synchronization mechanism not only has no obvious impact on the fault-tolerant performance,but also effectively reduces the total running time of tasks and optimizes the load of different cores in the system.Compared with the table scheduling algorithm,the DAG model-oriented fault-tolerant scheduling algorithm can effectively reduce the total running time when dealing with computing tasks with low balance coefficient when the same number of faults are randomly injected.