Towards distributed network classification for mobile ad hoc networks

Mobile ad hoc networks range from traditional MANETs where end-to-end paths exist from sources to destinations, to DTNs where no contemporaneous end-to-end paths exist and communication is achieved by the store, carry, and forward model of routing. Hence, nodes of these networks need to identify the level of connectivity of the network they belong to and classify it as a MANET or a DTN, in order to properly select appropriate protocols to achieve end-toend communication. What is more, since mobile ad hoc networks change over time and space, nodes need to periodically re-access their network classification to adapt to the always changing environment. Recently, there has been an effort to classify the various types of mobile ad hoc networks assuming there is a centralized authority that has complete knowledge of the network and its dynamics. In this paper we design distributed mechanisms for nodes to perform the above classification on the fly, based only on local information that they collect as they move and encounter other nodes. The mechanisms take advantage of a combination of measurements and analytical techniques. We investigate the accuracy of our mechanisms by comparing the network classification of the centralized authority to that of a node using our schemes.