Local versus global power adaptive broadcasting in ad hoc networks

In many ad hoc deployments, making broadcasting, a critical function component, power efficient is extremely important. Current power-adaptive approaches proposed for broadcasting, can be grouped into a) centralized or omniscient schemes, and b) decentralized or localized schemes. The latter assume only local neighborhood information. Due to the absence of global information, localized algorithms may not produce optimal or near-optimal solutions. On the other hand, global knowledge is not typically available to nodes and may be extremely expensive to disseminate. In this paper, we examine the importance of the lack of global information on the performance of localized approaches. Towards this, we perform extensive simulations and compare the performance of localized power adaptive broadcasting with the performance of the well-known Broadcast Incremental Power or BIP, an omniscient algorithm. We analyze the behaviors of the two protocols and identify the reasons for the differences in behavior. We observe that while global state does provide better performance in terms of energy efficiency, the localized scheme outperforms the global scheme in terms of the latency incurred in the broadcast. Based on the observed behavioral traits of the two protocols, we suggest changes by which the energy consumption with the localized scheme is reduced by as much as 20%, while incurring almost no penalty in terms of latency.