Design And Implementation Of Fault-Tolerant Parallel Algorithm For On-board Computer

On-Board Computer is the core of satellite service management and the satellite payload computing devices. Due to increasing of the application needed in space, the use of multi-CPU parallel has become the main development of On-Board Computer in the future. During operating in space, On-Board Computer equipment will remain in the radiation environment of charged particlesand, and the strong radiation, the high-temperature and the high-vacuum will make transient or permanent failures which affect the normal operation of satellite services, so the fault-tolerance technique is the resrarch hotport in the aera of On-Board Compute resrarching always. As an important aspect of Fault-tolerance, software fault-tolerance provides a flexible and effective way for improving the reliability of the system. In this paper, application-level software fault-tolerance is applied to tolerant the fault of hardware and improve system reliability basing on the analyzing the existing technology of fault-tolerance.Primary work in this thesis can be summarized as following:1,Facing the single-node fault of On-Board Parallel Computer, we propose the conception of the Fault-Tolerant Parallel Algorithm(FTPA),basing on the ideas of the parallel recomputing. Also we present the design process of the FTPA: Firstly, partition the program the section. Secondly, design the failure detection section and the data saving section. Finally, design the recovery section. By analyzing the structure characteristics and the application mode of On-Board Parallel Computer, we proposed two different implement methods of FTPA: distributing the work of the fault node and reload the replaced node.2,According to the design process of the FTPA and the two different implementation of fault-tolerance, we choose the NPB(NAS parallel Benchmark)which based on MPI(Message Passing Interface) communication interface and make the corresponding fault-tolerant design. At the same time we make the test in the PC of 4-processors and the result shows that the FTPA is feasible.3,The FTPA for On-Board Computer is a method of using software fault-tolerance to solve the hardware fault in application-level. Whether the design ideas and the implement methods of FTPA can be further works needed to meet the experiment and take the testing in the real hardware. We make plentiful study focusing on the communication theory and communication of the VxWorks operating system and make the test in functionality and performance in the VxWorks Simulator and the 4-processors prototype machine of On-Board Computer. The experiment results show that FTPA can improve system availability and reliability while improving application performance.