The Research Of Real-time Fault-Tolerant Mechanism In Distributed Real-time System DRTAS

An applied distributed real-time system must have high reliability. Being the most important way to provide high reliability, Fault-tolerance has been the necessary property of distributed real-time systems. However, current distributed real-time systems' fault-tolerance has some problems. There are too many restrictions, so the actual fault-tolerant effect is not good enough. They also often cause unnecessary resources being wasted and lack kernel-level fault-tolerance.Considering these problems, based on its structure, work flow, bottom real-time kernel and handling of real-time applications, distributed real-time system DRTAS provides two level real-time fault-tolerant mechanism. Top real-time fault-tolerance uses real-time fault-tolerant techniques as resume execution or re-execution and adopts the H-FTF scheduling method. And, based on time redundancy, bottom real-time fault-tolerance uses real-time fault-tolerant techniques as alternate version or re-execution and adopts the FT/FA scheduling method and the DSC-EDF algorithm. Moreover, two real-time fault-tolerant evaluations are added to the mechanism. The mechanism has efficient communication framework, uses precise time synchronization method and maintains system load balance. DRTAS's real-time fault-tolerant mechanism provides the system good fault-tolerant abilities and guarantees a reliable DRTAS.DRTAS's real-time fault-tolerant mechanism is implemented by using C and QT language in the Linux and RTAI platform. We test the mechanism's functions and performance. The functional tests validate that its scheduling and handling functions are realized and work successfully. The tests of performance indicate that periodic real-time tasks'event latency and period jitter are less than 30μs and that the system's real-time fault-tolerant cost is less than 85ms when the front-end node is under normal load and real-time fault-tolerant scheduler's timing is not more than 30ms.