I’ve been wondering what happened to the wireless mesh technology since I first read about it in 2006, when the One Laptop Per Child project was in its early stages and featured in Linux User and Developer. The idea was to mass produce a laptop at a cost of $100 per unit and distribute them to developing countries. One of the benefits would be that communities could set up their own wireless networks based on the 802.11s standard Mesh Routing Protocol, and maybe even connect to the Internet. While the OX-1 was finally deployed in reasonably large numbers, and was likely behind the creation of the netbook market, the mesh concept hasn’t got much publicity since.
One of the reasons we’re not seeing this technology in everyday life is the lack of the right software and hardware on the market. Curently the Marvel 8388 chipset is the only one that enables a system to function as a mesh relay independently of the CPU, which is important if there’s a limited number of nodes in a remote area, and there was some dispute over the availability of open source drivers and firmware for that. The exact details are posted somewhere on Cryptome.
A quick search of Wikipedia revealed there are around 70 protocols for mesh networking, none of which are yet common standards, but the OLPC project, the Open Technology Initiative, and other open source developers have settled with the Mesh Routing Protocol. At least that’s the situation as I understand it. Another disadvantage is no authentication is used, since none have been suggested that would be compatible with the design objectives. This might be a subject for my future dissertation.
To connect to a specific point in a mesh, an OLPC unit broadcasts a DCHP request and an RREP (request) signal on all (three) channels. The unit will receive one or more responses, and determines which nodes to send its traffic through, based on the efficiency and signal strengths of all available links. More details on how that works are available on its networking wiki.
As we can see from the diagram below, an OLPC can function as a gateway to the Internet or as a basic wireless repeater, and only one node needs to be conncted directly to the Internet to provide the whole network with access. It should be possible for a single router to link several meshes to the Internet, if that’s needed.
The main difference between this and conventional networks is the absence of a single data bus, and the routing protocol, used for determining the most efficient route between two nodes, also handles the load balancing within the mesh.
This is suited to situations where OLPC units are distributed in large numbers in locations where there’s no permanent infrastructure, telephone lines or electricity. Local Area Networks are sometimes divided into segments connected by switches. In the case of mesh networks, each channel is a separate collision domain.
In practice, conventional servers are being installed at some locations to handle the email and Internet access for the mesh networks.
Repressive governments will find it very difficult to bring down a mesh network without sophisticated electronic warfare equipment capable of flooding multiple frequencies over a wide geographic area, and this is the main reason the US government’s funding a project, known as ‘Internet in a suitcase’, to deploy mesh networks in Afghanistan alongside a protected cellular network and several gateways near the border to provide Internet access.
Developed by hackers, information security experts and commercial organisations, such a network not only gives freedom of expression to the Afghans, it can be used to gather intelligence and circulate propaganda far more effectively than a one-way system like Voice of America. There are numerous other benefits in terms of social development.
Another project separate from the ‘Internet in a suitcase’ is the FabFi network that was field tested in another part of Afghanistan and Kenya several years before. Using the Fab Lab system, locals were able to build their own antennae and parts of the infrastructure from surplus materials. They were then able to mount this onto commercial hardware so they could be wirelessly interconnected. The end result is a wide area wireless backbone for local OLPC units. Some further development was required to provide the nodes with reliable access to the Internet and ensure the available bandwidth is allocated fairly to all users. This enabled one of the groups in Kenya to set up a co-operative ISP.
By all accounts, the FabFi network in Afghanistan provides pretty solid Internet access in that part of the country capable of delivering 11.5Mbps over the maximum effective range of 6KM, as of early 2009. Work is still ongoing, and the developments can be read about on the FabFi project’s blog.