Models of 802.11 multi-hop networks: Theoretical insights and experimental validation

Abstract

Wireless Multi-Hop CSMA/CA Networks are challenging to analyze. On the one hand, their dynamics are complex and rather subtle effects may severely affect their performance. Yet, understanding these effects is critical to operate upper layer protocols, such as TCP/IP. On the other hand, their models tend to be very complex in order to reproduce all the features of the protocol. As a result, they do not convey much insight into the essential features. We review two models of 802.11 protocols, which are simple enough to first explain why a trade-off needs to be found between fairness and spatial reuse (throughput) in saturated wireless networks (where all nodes have packets to transmit to their neighbors); and then to explain why non-saturated networks (where only some nodes, the sources, have packets to transmit to their destinations in a multi-hop fashion) that are more than 3 hops longs suffer from instability.We confront both models either to realistic simulations in ns-2 or to experiments with a testbed deployed at EPFL. We find that the predictions of both models help us understand the performance of the 802.11 protocol, and provide hints about the changes that need to be brought to the protocol.