On Adaptive Fault Diagnosis for Multiprocessor Systems

Abstract. We first consider adaptive serial diagnosis for multiprocessor systems. We present an adaptive diagnosis algorithm using N + t − 1 tests, which is the smallest possible number, for an N-processor system modeled by a (2t − 1)-connected graph with at most t faulty processors. We also present an adaptive diagnosis algorithm using minimum number of tests for a system modeled by cube-connected cycles. We consider adaptive parallel diagnosis as well. We show that for adaptive parallel diagnosis of an N-processor system modeled by a hypercube, three testing rounds are necessary and sufficient if the number of faulty processors is at most log N − log(log N − log log N +4) +2. We also show that three testing rounds are necessary and sufficient for adaptive parallel diagnosis of a system modeled by cube-connected cycles of dimension greater than three.

• Type: Conference
• Year: 2001
• Where: ISAAC
• Authors: Kumiko Nomura, Toshinori Yamada, Shuichi Ueno
• Keywords: Adaptive Diagnosis Algorithm, Adaptive Parallel Diagnosis, Adaptive Serial Diagnosis, Algorithms, ISAAC 2001