The Research Of The Real-time Fault-tolerant Scheduling Algorithm For Hybrid Tasks In Distributed System

Distributed real-time system is used as the core control component for safety-critical applications, and is widely used in construction machinery automation, flight control systems, and industrial control distributed environment, become decisive factors affect system security and reliability. Fault-tolerant technology as a guarantee of real-time systems reliability has become an important research field of real-time system. Distributed real-time system must not only meet the time requirements of periodic tasks in data acquisition and transmission, but also to ensure that the time constraints of aperiodic tasks in control and alarm.This paper summarizes the distributed real-time systems task scheduling features, propose real-time fault-tolerant scheduling algorithm for hybrid task.The main work is as follows:(1) Introduced the system models and the characteristics of distributed systems, fault-tolerant systems and real-time systems. On this basis, reviewed the development of distributed real-time fault-tolerant scheduling algorithms.(2) Proposed a Hybrid Tasks-Space and Time Redundancy scheduling algorithm (HT-STR), which can schedule periodic and aperiodic tasks with fault-tolerant requirements, and aperiodic tasks with non fault-tolerant requirements aperiodic tasks. There are two versions (major version and minor version) for each task with fault tolerance requirements. Periodic tasks with fault-tolerant requirements are divided into high-frequency task set and low-frequency task set. We can use space redundancy strategy for high-frequency tasks and time redundancy strategy for low-frequency tasks. Simulation results show that in the process of scheduling, the algorithm can effectively shorten the processor communication time and make the system more balanced load.There is higher resource utilization and no time overlap.(3) Came up with an algorithm based on feedback control of real-time fault-tolerant scheduling. It introduced the feedback mechanism into the HT-STR algorithm. The algorithm is to set the reference input value according to the foreknowledge of the processor utilization scope for each processor, then compare the processor utilization with the reference input value to get control input value. If the control input is greater than zero, processor overload, cancel the non-critical portions of periodic tasks and the minor version of the tasks with fault-tolerant requirements, so that the load can balance. If the control input is less than zero, processor underused, the task allocator will select a number of pending tasks, assigned to the current processor to improve processor utilization. Simulation results show that the algorithm enables the processor utilization to remain in the vicinity of the reference value and prevent the loss of the task, so that it can ensure the real-time and reliability of the task.