Improving Throughput in Multihop Wireless Networks

Abstract

One of the main characteristics of wireless ad hoc networks is their node-centric broadcast nature of communication, leading to interferences and spatial contention between adjacent wireless links. Due to such interferences, pessimistic concerns have been recently raised with respect to the decreasing network capacity in wireless ad hoc networks when the number of nodes scales to several orders of magnitude higher. Such studies assume uniformly distributed nodes in the network and randomized traffic patterns. In this paper, we argue that in all cases of end-to-end data communications-including one-to-k unicast and multicast data dissemination as well as k-to-one data aggregation-the maximum achievable end-to-end data throughput (measured on the sources) heavily depends on the strategy of arranging the topology of transmission between sources and destinations, as well as possible per-node operations such as coding. An optimal strategy achieves better end-to-end throughput than an arbitrary one. We present theoretical studies and critical insights with respect to how these strategies may be designed so that end-to-end throughput may be increased. After all, under all circumstances-in either a lightly loaded or a congested network-increasing end-to-end throughput from its baseline is always beneficial to applications using ad hoc networks to communicate.