Design and Evaluation of Online Fault Diagnosis Protocols forwireless Networks

Any node in a network, or a component of it may fail and show undesirable behavior due to physical defects, imperfections, or hardware and/or software related glitches. Presence of faulty hosts in the network affects the computational efficiency, and quality of service (QoS). This calls for the development of efficient fault diagnosis protocols to detect and handle faulty hosts. Fault diagnosis protocols designed for wired networks cannot directly be propagated to wireless networks, due to difference in characteristics, and requirements. This thesis work unravels system level fault diagnosis protocols for wireless networks, particularly for Mobile ad hoc Networks (MANETs), and Wireless Sensor Networks (WSNs), considering faults based on their persistence (permanent, intermittent, and transient), and node mobility. Based on the comparisons of outcomes of the same tasks (comparison model ), a distributed diagnosis protocol has been proposed for static topology MANETs, where a node requires to respond to only one test request from its neighbors, that reduces the communication complexity of the diagnosis process. A novel approach to handle more intractable intermittent faults in dynamic topology MANETs is also discussed.Based on the spatial correlation of sensor measurements, a distributed fault diagnosis protocol is developed to classify the nodes to be fault-free, permanently faulty, or intermittently faulty, in WSNs. The nodes affected by transient faults are often considered fault-free, and should not be isolated from the network. Keeping this objective in mind, we have developed a diagnosis algorithm for WSNs to discriminate transient faults from intermittent and permanent faults. After each node finds the status of all 1-hop neighbors (local diagnostic view), these views are disseminated among the fault-free nodes to deduce the fault status of all nodes in the network (global diagnostic view). A spanning tree based dissemination strategy is adopted, instead of conventional flooding, to have less communication complexity. Analytically, the proposed protocols are shown to be correct, and complete. The protocols are implemented using INET-20111118 (for MANETs) and Castalia-3.2 (forWSNs) on OMNeT++ 4.2 platform. The obtained simulation results for accuracy and false alarm rate vouch the feasibility and efficiency of the proposed algorithms over existing landmark protocols