Each node in the Mesh network doesn't just send and receive data for itself, it also routes messages on for others. This might seem retrograde at first, as each node ends up sharing bandwidth with traffic in which it has no interest. There are also many practical difficulties in making a Mesh network efficient and reliable: routing a packet is much harder if the links between it and its destination are liable to change at a second's notice. However, the practical benefits of a Mesh network are potentially huge.
Take bandwidth. In a traditional wireless network, there's a fixed bandwidth shared between all stations within range of an access point. The more stations share that space, the less bandwidth per station is available. With Mesh networks there is no need to use the access point: each station can reduce its power until it is in contact only with those stations closest to it. Thus, other stations nearby are then free to reuse the frequencies for connections to their own neighbours and the total bandwidth in the area goes up thanks to spatial and frequency reuse. The more stations you have, the more bandwidth becomes available.
This also has positive implications for reliability and efficiency. With a dense Mesh of network nodes, failure of one merely means that packets moving across the network have to take an alternate route -- compare this to the situation with current wireless LANs where an access point goes down. Also, a Mesh network lends itself to load balancing -- each node can choose routes according to which is least congested, thus making best use of the bandwidth and allowing quality of service considerations to be a part of the routing process.
There are two main kinds of Mesh networks: ones in which every station is fixed, and ones where the stations are free to move around. It might seem that the first variety is simpler than the second, as once the routing map has been set up the system will be similar to a wired network. The second type would appear to be far more complicated, as stations can appear, disappear and move in and out of range of each other all the time.
The fixed model works well for networks distributing data between buildings. However, it breaks down at smaller levels, in the office or home. At the microwave frequencies used, any change in the physical environment -- including people walking around, furniture moving and so on -- can block one radio path or create another through reflections. In effect, everything is moving around the network map all the time, and any routing protocol has to accept and work with that. A simplistic solution won't work.
Mesh is a new concept in commercial terms -- the military and others have been researching it for a while -- and standards are thin on the ground. One of the most advanced is 802.16a, which has grown out of the 802.16 fixed wireless broadband standard. 802.16 is designed for multi-megabit links from 10 to 66 GHz in star configurations -- the sort of broadband to the home or office that uses a central fixed distribution point feeding many small fixed transceivers. Mesh was seen as a way to fill in gaps in that model of coverage. 802.16a added that in conjunction with lower powers and lower frequencies than 802.16.
Many questions remain about how the 802.16 family of standards and existing wireless networks will co-exist: there are strong rumours that Intel, among others, is developing systems with aspects of both. Intel has already demonstrated Mesh networks running on top of standard 802.11b interface cards. One model has 802.16 connecting islands of 802.11; another has 802.16a subnets connected to a building-wide 802.11. One potentially dynamite combination would be 802.16a and ultrawideband (UWB) radios: UWB is a new carrier-less radio technology
whose very short range and very high bandwidth, make it ideal for dense Meshes. This model predicts £10 nodes every five metres or so throughout an office or home, providing multiple 400 megabit/second bandwidth connections.
The three main areas needing work are security, manageability and routing. Security and manageability are two aspects of the same problem: a Mesh network relies on having as many nodes as possible co-operating, but you don't want an outside agency stealing all your bandwidth or frequencies. You do want to co-operatively share neighbouring networks: you don't want your packets being vulnerable to outside interference. You want to be able to increase your network's capacity by adding blocks of new nodes securely: you don't want to have to spend all your time configuring or maintaining them. Routing is still being experimented with: here the tradeoff is that routing is most efficient when each node knows all about the network around it, but with a dynamically changing network you end up using all your bandwidth telling everyone about the changes. Various compromises exist, and are being ascertained for robustness, efficiency, configurability, latency and so on.
In the future, Mesh networks may extend outdoors. Two new standard efforts -- 802.16e and 802.20 -- are concentrating on mobile broadband wireless networking, in other words the megabit-upward connectivity on the move that will characterise 4G mobile communications. Whatever happens, Mesh is one technology to watch.