The multi-level paradigm for distributed fault detection in networks with unreliable processors

In this paper, we study the effectiveness of the multilevel paradigm in considerably reducing the diagnosis latency of distributed algorithms for fault detection in networks with unreliable processors. This study is based on our work in [6] that generalizes the algorithm in [3]. In 1967, Preparata, Metze and Chien [I] proposed a model called the PMC model and a framework, called System-Level Diagnosis, for diagnosing faults in a network of processors. The PMC model allows faulty processors also to test. Thus some of the results could be unreliable. In the more than three decades following this pioneering work, several issues arising from the application of this framework have been investigated and resolved. Diagnosis algorithms based on the PMC model are assumed to be executed on a single highly reliable supervisory processor. A single supervisory processor is a bottleneck in a system with a large number of processing elements. Distributed diagnosis algorithms exploiting the inherent ...

• Type: Conference
• Year: 2003
• Where: ISCAS
• Authors: Krishnaiyan Thulasiraman, Ming-Shan Su, V. Goel
• Keywords: Algorithm, Diagnosis Latency, ISCAS 2003, PMC Model