Adaptive Diagnosis of Variants of the Hypercube

Abstract

System-level fault diagnosis deals with the problem of identifying faulty nodes (processors) in a multiprocessor system. Each node is faulty or fault-free, and it can test other nodes in the system, and outputs the test results. The test result from a node is reliable if the node is fault-free, but the result is unreliable if it is faulty. In this paper, we prove that four variants of the hypercube: the crossed cube, the twisted cube, the Mobius cube, and the enhanced cube, are adaptively diagnosed using at most 4 parallel testing rounds, with at most n faulty nodes (for the enhanced cube, with at most n + 1 faulty nodes), where each processor participates in at most one test in each round. Furthermore, we propose another diagnosis algorithm for the n-dimensional enhanced cube with at most n + 1 faulty nodes, and show that it is adaptively diagnosed with at most 5 rounds in the worst case, but with at most 3 rounds if the number of existing faulty nodes is at most n - ⌊log (n + 1)⌋.