The Optimized Link State Routing Protocol Evaluation through Experiments and Simulation

In this paper, we describe the Optimized Link State Routing Protocol (OLSR) [1] for Mobile Ad-hoc NETworks (MANETs) and the evaluation of this protocol through experiments and simulations. In particular, we emphasize the practical tests and intensive simulations, which have been used in guiding and evaluating the design of the protocol, and which have been a key to identifying both problems and solutions. OLSR is a proactive link-state routing protocol, employing periodic message exchange for updating topological information in each node in the network. I.e. topological information is flooded to all nodes in the network. Conceptually, OLSR contains three elements: Mechanisms for neighbor sensing based on periodic exchange of HELLO messages within a node’s neighborhood. Generic mechanisms for efficient flooding of control traffic into the network employing the concept of multipoint relays (MPRs) [5] for a significant reduction of duplicate retransmissions during the flooding process. And a specification of a set of control-messages providing each node with sufficient topological information to be able to compute an optimal route to each destination in the network using any shortest-path algorithm. Experimental work, running a test-network of laptops with IEEE 802.11 wireless cards, revealed interesting properties. While the protocol, as originally specified, works quite well, it was found, that enforcing “jitter” on the interval between the periodic exchange of control messages in OLSR and piggybacking said control messages into a single packet, significantly reduced the number of messages lost due to collisions. It was also observed, that under certain conditions a “naive” neighbor sensing mechanism was insufficient: a bad link between two nodes (e.g. when two nodes are on the edge of radio range) might on occasion transmit a HELLO message in both directions (hence enabling the link for routing), while not being able to sustain continuous traffic. This would result in “route-flapping” and temporary loss of connectivity. With the experimental results as basis, we have been deploying simulations to reveal the impact of the various algorithmic improvements, described above.