Using temporal correlation for fault localization in dynamically changing networks

A mobile ad hoc network creates a dynamic environment where node mobility can cause periodic changes in routes. Most existing fault localization algorithms assume availability of a complete and/or deterministic dependency model. Such assumptions cannot be made in the dynamically changing networks. This paper is aimed at developing a fault diagnosis architecture and algorithm to address the issue of dynamically changing dependencies in networks. We propose an architecture to capture the changes in dependencies and introduce a temporal correlation algorithm to perform fault diagnosis with the dynamically changing dependency information. We present an experimental evaluation of our work through simulation results using Qualnet. Copyright © 2007 John Wiley & Sons, Ltd.