An Algorithmic Approach to Conditional-Fault Local Diagnosis of Regular Multiprocessor Interconnected Systems under the PMC Model

• Published in: IEEE transactions on computers Vol. 62; no. 3; pp. 439 - 451
• Main Authors: Cheng-Kuan Lin, Tzu-Liang Kung, Tan, J. J. M.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.03.2013; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: algorithm; Algorithms; Architecture (computers); Computer simulation; diagnosability; Diagnosis; Diagnostic systems; Fault diagnosis; Faults; graph; Hypercubes; Interconnected systems; Multiprocessing systems; Multiprocessor; PMC model; Processors; Program processors; reliability; Studies
• ISSN: 0018-9340; 1557-9956
• DOI: 10.1109/TC.2011.249

System-level diagnosis is a crucial subject for maintaining the reliability of multiprocessor interconnected systems. Consider a system composed of N independent processors, each of which tests a subset of the others. Under the PMC diagnosis model, Dahbura and Masson proposed an O(N 2.5 ) algorithm to identify the set of faulty processors in a t-diagnosable system, in which at most t processors are permanently faulty. In this paper, we establish some sufficient conditions so that a t-regular system can be conditionally (2t-1)-diagnosable, provided every fault-free processor has at least one fault-free neighbor. Because any t-regular system is no more than t-diagnosable, the approached diagnostic capability is nearly double the classical one-step diagnosability. Furthermore, a correct and complete method is given which exploits these conditions and the presented branch-of-tree architecture to determine the fault status of any single processor. The proposed method has time complexity O(t 2 ), and thus can diagnose the whole system in time O(t 2 N). In short, not only could the diagnostic capability be proved theoretically, but also it is feasible from an algorithmic perspective.