Performance Evaluation of Broadcasting Protocols for Ad Hoc and Sensor Networks

In ad hoc and sensor networks, the simplest and most widely used approach to broadcast is blind flooding, which lets every node in the network to rebroadcast a receiving packet to all its neighbors. This causes redundancy of broadcast packets and results in unnecessary collisions and bandwidth waste. To overcome these problems, a number of research groups have proposed more efficient broadcasting schemes with the goal of minimizing the retransmissions, while still guaranteeing that a broadcast packet is delivered to all the nodes in the network. Multipoint relay (MPR) and dominating set (DS) based broadcasting schemes can effectively improve the broadcasting efficiency while providing reliable broadcasting. The neighbor elimination scheme (NES) can improve any broadcasting protocol as an added feature. In this paper, we evaluate the performance of MPR (source dependent), MPR-DS (source-independent MPR), and DS based broadcasting protocols. We add NES to these three schemes separately and evaluate the performance of the resulted protocols. In our experiments, we use the random unit graphs to model the ad hoc and sensor networks. Each of the studied protocols has scenarios under which it has the best performance. Our experiments demonstrate that, without applying neighbor elimination scheme, MPR based protocol requires fewest retransmissions (however, each retransmission is with a longer message including list of forwarding neighbors). DS and MPR-DS schemes benefit significantly from the neighbor elimination technique in terms of the ratio of re-broadcasting nodes and the message redundancy on both transmitting and nontransmitting nodes, while MPR benefits marginally. After adding the neighbor elimination scheme, three new protocols behave almost equally well in terms of rebroadcast message counts. MPR-NES method is narrowly the best when the message that is broadcasted is a very large one, and the network is dynamic. MPR-DS-NES is narrowly the best when the broadcast message is not very large, and the network is stable (this method requires the third round of preprocessing HELLO messages). Overall, DSNES appears to be the most robust, taking all measurements and parameters into acount, because it remains competitive under all scenarios, and has significant advantages over MPR-DS-NES in dynamic scenarios, and significant advantages over MPR-NES when the broadcast message is not very large, because MPR has overhead in packet lengths.