Fault-tolerance method of large-scale heterogeneous parallel computing

Abstract

The invention provides a fault-tolerance method of large-scale heterogeneous parallel computing. The method includes treating each core computing module of each time step as follows: contents in computing digit groups of a computing subject are assigned to backup digit groups and computing for completing the core computing modules is performed; wherein the computing for completing the core computing modules includes that the number of available processor cores is computed to obtain the number of a first processor cores; the available processor cores perform parallel computing on the core computing modules; the number of the available processor cores is computed for a second time to obtain the number of a second processor cores; and the number of the first processor cores and the number of the second processor cores are compared, if the number of the first processor cores is smaller than that of the second processor cores, contents of backup digit groups are assigned to the computing digit groups, and computing for completing the core computing modules is completed for a second time until the number of the first processor cores is consistent with that of the second processor cores. According to the fault-tolerance method of large-scale heterogeneous parallel computing, computing resources can be fully used, fault recovery time is reduced, and the reliability of parallel computing is improved.

Description

technical field [0001] The invention relates to the field of computers, in particular to a fault-tolerant method for large-scale heterogeneous parallel computing. Background technique [0002] Large-scale heterogeneous high-performance computer systems are an important development direction for extremely large-scale parallel computing in the future. Compared with traditional single-core / multi-core processor computer systems, large-scale heterogeneous high-performance computer systems are based on heterogeneous processors. The number of processor cores has increased dramatically, and the system architecture and memory access methods have undergone major changes. In the environment of large-scale heterogeneous computer systems, how to ensure the reliability and stability of large-scale parallel computing is a key issue, and the fault-tolerant mechanism and intermittent processing mechanism at the parallel algorithm level are one of the key technologies. Efficient algorithm-le...

AI Technical Summary

Problems solved by technology

[0003] At present, in terms of fault tolerance mechanism and intermittent processing mechanism at the parallel algorithm level, the parallel computing software in the main application fields has designed and developed a large-scale parallel algorithm with arbitrarily variable parallel scale retention and recovery functions, which can ensure that the algorithm is at the MPI level (“ Message Passing Interface", but at the many-core parallel level, due to the particularity and complexity of architectures such as GPU ("Graphics Processing Unit", Graphic Processing Unit) and Cell processors, there are few The application program considers the implementation of the fault-tolerant function at the many-core level. During the calculation process, it cannot automatically detect the hardware failure of the large-scale heterogeneous computer system at the processor core level. It can only judge whether the calculation is normal and reliable from the final calculation result. Reliability and Stability of Scale Heterogeneous Parallel Computing

For some medium and large-scale many-core parallel projects with long computing time, it often happens that the project is suspended, and manual intervention is required to resubmit

[0004] Taking the application field of numerical simulation of the entire watershed of aerospace vehicles as an example, according to the currently available literature, the existing heterogeneous many-core parallel only realizes the recording of intermediate calculation results, that is, realizes the general retention and recovery function, and does not take into account the many-core level. The fault-tolerant function is realized. During the calculation process, the hardware failure of the large-scale heterogeneous computer system at the processor core level cannot be automatically detected. It can only be judged from the final calculation result whether the calculation is normal and reliable.

Images