Research Of Rollback Recovery Mechanism Based On Domestic-processor Computer System

Domestic processors develop rapidly and have been widely used in manyfields. But in the aviation, military industry, finance and other important fields,domestic processors couldn’t meet the requirements of high reliability. Research offault tolerant technology in different levels has important meanings to our countryin these important areas. It is important for breaking the foreign monopoly.Research of rollback recovery mechanism based on computer system with domesticprocessor chip is aimed at improving the fault-tolerant performance of domesticprocessor system in the operating system level, it is of great significance.In this paper, through the study of the loongson3a processor and the MIPSarchitecture, analyzed the characteristics of the loongson system platform. On thisbasis, I studied the rollback recovery mechanism and Linux process, and determineto keep which information when set checkpoint.Based on research of process management and operation mechanism, thispaper designs a kernel-level checkpoint rollback recovery mechanism. Save theprocess context information in kernel, when fault occurs, rollback to recovery.Kernel-level checkpoint is transparent to user; it doesn’t need to modify the user’scode. Kernel module can load and unload flexibly, and it needn’t change theoriginal system. This paper expounds the realization method of each module,mainly including: fault daemon module, checkpoint information saving module,rollback recovery module, file processing module and so on.Finally I test the basic application of rollback recovery system, and analyzedthe factors of affecting the system in time and space costs through repeatedlyexperiment. Through changing the strategy of set checkpoint, reduce checkpointtime overhead. Improving the method of saving the open files reduced thecheckpoint space overhead. In all, this rollback recovery mechanism improves thefault tolerance of system, at the same time it has a better usability.