Open AccessFault‐tolerance in wireless ad hoc networks: bi‐connectivity through movement of removable nodes
Author(s) -
Yan Zhongjiang,
Chang Yilin,
Jiang Hai,
Shen Zhong
Publication year - 2011
Publication title -
wireless communications and mobile computing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.42
H-Index - 64
eISSN - 1530-8677
pISSN - 1530-8669
DOI - 10.1002/wcm.1164
Subject(s) - computer science , computer network , node (physics) , wireless ad hoc network , ring network , mobile ad hoc network , wireless mesh network , wireless network , distributed computing , wireless , network packet , network topology , engineering , telecommunications , structural engineering
For a wireless ad hoc network to achieve fault‐tolerance, it is desired that the network is bi‐connected. This means that each pair of nodes in the network have at least two node‐disjoint paths between them, and thus, failure at any single node does not partition the network. In other words, in a bi‐connected network, there is no cut‐node (defined as a node such that the removal of it partitions the network). To make a connected but not bi‐connected network become bi‐connected, actions should be taken such that all cut‐nodes become non‐cut‐nodes. In this research, we propose to deal with cut‐nodes from a new perspective. Specifically, we first introduce a new concept of removable node , defined as a non‐cut‐node such that the removal of it does not generate any new cut‐node in the network. Then, we propose to move a removable node to a new location around a cut‐node. In this way, the cut‐node becomes a non‐cut‐node, that is, the failure of it does not partition the network anymore. Algorithms are provided (i) to identify removable nodes; (ii) to match cut‐nodes with a feasible set of removable nodes, in which all nodes can be simultaneously removed from the network without generating any new cut‐node in the network; and (iii) to derive the final location of a removable node such that its movement distance is the shortest and the associated cut‐node becomes a non‐cut‐node. The proposed algorithms do not guarantee the final bi‐connectivity but have the merits of a large success rate (almost 100% in the simulation), a small number of moved nodes, and a short total movement distance. In addition, the proposed algorithms are shown to be effective even when there are a large portion of fixed nodes in the network. Copyright © 2011 John Wiley & Sons, Ltd.