Component Diagnosis Strategy of Star Graphs Interconnection Networks

The growing demand for high-performance computing and the acceleration of information processing have brought multiprocessor systems into the era of E-class computing. The reliability of interconnection networks built on multiprocessor systems is facing severe challenges with the rapid growth of the networks' scale. For example, a large-scale processor failure may disconnect the entire network and result in the appearance of many different components. Rapid fault diagnosis has great advantages in industry, especially in real-time systems, which means that rapid diagnosis of fault processors is particularly important. However, the fault diagnosis capability in a network is closely related to the amount of components that the network can tolerate in its application. The diagnosis method of faulty processors, which cause many components is called component diagnosis. In particular, ct(g)(G) refers to the maximum number of faulty processors meeting the g -component condition that can be diagnosed in network G under certain system-level diagnostic model. In this article, based on the indistinguishability of the constructed set and linear multiple faults analysis technology, we propose the 2, 3-component diagnosabilities of star graph network S-n under P-M-C-M. Moreover, we propose a novel g -component t -diagnosable algorithm FCFDSn innovatively to diagnose all faulty processors, and we implement the algorithm FCFDSn on both synthetic data and real data. Furthermore, we verify the availability/efficiency of algorithm FCFDSn in terms of true positive rate, true negative rate, and accuracy rate.