Study Of Distributed Fault-tolerant Computer Redundancy Management Techniques

The development of computer brings many advantages to people's life and human being become more and more reliant to computers. The software and hardware of a computer system would create certain fault inevitably. Moreover, this kind of fault may cause incalculable loss. Hence, enhancing the reliability of computer is another goal in the process of computer development. Fault tolerance is an effective method to enhance the reliability of computers.Under the background of flight control computer, three aspects, synchronization, voting and reliability analysis, of Triple Modular Redundancy (TMR) fault-tolerant computer system are discussed.First, synchronization warrants every module can work in step. In this way, the voting, changing between modules and reconstructing of system would make sense. Based on redundancy management of triple modular redundancy fault-tolerant computer system, a task synchronization algorithm is developed. The result of emluator indicates that this algorithm can synchronize the datum of modules and provide imputing datum for voting very well.Second, voting is a mechanism of fault tolerant. It can mask some faults of computer system. Majority voting and plurality voting are two voting algorithms. In order to get an appropriate outputting data of these voting algorithms, a clustering algorithm is designed. This algorithm divides N modules into classes, producing a class that includes more elements. The result shows that clustering can work well, but is reliant to threshold gaved by system designer.Last, reliability analysis would provide academic sustainability for designing and building a fault-tolerant computer system. On the basis of a Fault Tree, partition set and minimum partition set are used to calculate the reliability. The result of analysis indicates that the fault tolerance technology can enhance the reliability of computer system.