Research And Implementation Of High Reliable Real-Time Software Architecture Based On Multi-Processors

The continuous improvement of semiconductor technology promotes the rapid development of multiprocessor technology.The increase of transistor density and the decrease of working voltage not only improve the performance of multiprocessor and reduce the power consumption,but also make it more vulnerable to the interference of external environmental factors,such as transient fault and other abnormal conditions.In order to effectively manage the computing resources of the multi-core processor as well as achieve fault tolerance and high availability,a fault-tolerant software architecture for real-time software is designed and implemented based on multi-core processor in this article.A task migration mechanism for redundant fault tolerant tasks is proposed,and a checkpoint synchronization mechanism for load sensing is further proposed.This article proposes a low overhead task migration method for redundant fault-tolerant tasks.Based on the condition that there are copies of the same task on different processing nodes and the redundant copies of tasks in the synchronous state can be replaced with each other,by acquiring communication overhead between processing nodes dynamically or statically,we can replace the migration source node of transferring task state by other low overhead transmission source nodes in the process of task migration so as to reduce its overhead.Based on the task migration method,a load aware checkpoint synchronization method is proposed for the limited computing resources in the embedded environment.This method monitors the execution progress of redundant tasks on different processing nodes by checkpoint index,and takes this index as the load state of processing nodes.When the index of a redundant task copy lags behind the other two copies,the checkpointing information of the other two copies is used to update and replace the lagging copy,so as to realize heavy load sensing and task acceleration and task reduction.In this paper,a fault tolerant software architecture based on multi-core processor is designed and implemented.It can shield transient faults by task redundancy and shield permanent faults through core redundancy.Combined with the needs of real-time software,the roll-forward error correction scheme implemented by three modular redundancy and checkpointing mechanism improves the real-time and reliability of the whole system.Finally,this article tests the system by the method of software fault injection,and the test shows that the software architecture can effectively detect and deal with transient faults.Compared with the common task migration mechanism,the implementation of redundant task migration method has lower overhead.The implementation of load aware checkpointing mechanism has better performance in the case of limited resources.