Using local conditions to reduce control overhead

In this paper, we replace the periodic transmission of control information by transmissions that depend on local network conditions. As a case study, we consider the proactive link state routing protocol OLSR running in a wireless network without infrastructure. Each node maintains a time series on its betweenness, a metric widely used in social network analysis. We interpret an anomaly in a node’s time series as a change in its role and use it to trigger the transmission of link state. In order to ensure throughput does not degrade by the use of dated information, we use a keep-alive timer whose interval takes into account other local conditions including the node speed, packet arrival rate, and number of flows served. We also measure the number of local link breaks and use it to trigger the transmission of neighbour information. ns-2 simulations comparing our proposed A+-OLSR to OLSR, A-OLSR, and Adaptive OLSR show a statistically significant increase in throughput and a decrease in control overhead. Our evaluation also considers packet losses and the use of CRAWDAD wireless traces to drive node movement.