A Feature-Oriented Fault Diagnosis Agreement Protocol in Distributed Systems

Understanding the fault-tolerance of distributed systems is crucial for achieving reliability. One of the most important issues surrounding fault-tolerance is the Fault Diagnosis Agreement (FDA) problem. The purpose of FDA is to help each fault-free processor detecting/locating a common set of faulty processors. In general, FDA protocols need ⎣ ( n -1)/3 ⎦ +2 rounds of message exchange to detect or locate faulty components, regardless of the presence or absence of faulty processors. However, the number of messages produced results in a large protocol overhead. To solve the FDA problems efficiently, a novel concept of feature value is proposed in our algorithm to reach an agreement using the minimum constant number of rounds characterized by the presence of dual failure characteristics of processors. In addition, the proposed protocol can detect/locate the maximum possible number of faulty processors in a network.