The ts-Diagnosability of Networks via Component Connectivity

With the growing role of multiprocessor systems in big data, artificial intelligence, as well as cloud computing and high-performance computing, the expansion in system scale and complexity has inevitably led to an increase in processor failures (or faults). To enhance the system’s fault diagnosis capability, a novel approach termed ts-diagnosis has been proposed timely. In this article, we delve into the relationship between a general network’s component connectivity and ts-diagnosability under the MM* diagnostic model, and subsequently apply the metric to a variety of networks, including bubble sort graphs, complete cubic networks, hierarchical hypercubes, and generalized exchanged hypercubes. To detect all faulty nodes, we propose the largest connected component under MM* model (LCC-MM*) algorithm along with the analysis of time complexity. In addition, we evaluate the ts-diagnosability of a variety of networks, accompanied by contrasting ts-diagnosability with other conditional diagnosabilities under the MM* diagnostic model. Meanwhile, we conduct experiments on real data to assess the effectiveness and performance of the LCC-MM* algorithm.