Network management and system-level diagnosis

Fault management, which consists of fault detection, diagnosis, and recovery is one of the key goals of network management. We consider an application of system-level diagnosis concepts to the problem of fault diagnosis in networks. Since the primary function of the nodes in a communication network is to route messages, the diagnosis is done with respect to the ability of the nodes to accomplish routing correctly. We propose a basic test model in which a node is tested with the help of two other nodes (a tester and a helper node) that check the ability of the tested node to correctly route the messages between the tester and the helper nodes. Necessary conditions are derived for a network to be t-diagnosable under this model. The proposed model does not require tests with perfect coverage.