Wireless Mesh Networks (WMNs) have been the topic of much research over the past few years. Challenges inherent to transmitting data over a wireless medium have caused researchers to rethink many commonly held assumptions in how to implement a network. For example, the Hidden and Exposed Terminal problems are accountable for much of the collision-based loss in a wireless network. Also, the design of the 802.11 MAC is not optimal for multiple nodes to send concurrently. At the network layer, route selection is different because it depends on the quality of the signal between any two nodes, and the ability of nodes to back off while others transmit. Because most of the loss in WMNs does not come from queue overflow, important assumptions behind TCP break down.
One paper evaluates Roofnet, a test bed in Cambridge, MA. It shows that, unlike wired networks, when wireless nodes are denser, throughput often increases. It also shows that several nodes are responsible for most of the network's success. Another paper uses a simulations of a wireless backhaul network to show that there is significant unfairness when multi-hop flows compete against single-hop flows. One problem is that the 802.11 MAC has a backoff mechanism that makes it difficult for a node to transmit to a medium when another node is already transmitting. Another issue is that the hidden terminal problem causes a lot of loss from collisions.
In short, WMNs are a new and interesting, unsolved research topic. I'm interested to see what conclusions will be made after several more years of research in Wireless Mesh Networks.
No comments:
Post a Comment