A Fine Mesh
By Kevin Werbach, Tue Mar 30 10:30:00 GMT 2004

Meshes could be the future of wireless... if anyone could make them work in a real-world deployment. Fortunately, there signs that mesh networking is finally set to achieve its potential.

It always sounded like a good idea: wireless devices that talk to each other instead of to central towers. Just like the routers the make up the Internet, nodes in these "mesh networks" can self-organize and become more robust as network grows. After several embarrassing failures, a number of companies are commercializing mesh networking systems for a variety of applications.

In a mesh network, the devices cooperate. Each wireless receiver can also be a relay, passing traffic on to other nodes. Because every device doesn't need a direct connection back to a central transmitter, the network uses wireless capacity more efficiently and can operate in places that otherwise couldn't be served.

If there is a hill between my house and the transmission tower, for example, I may not get a good enough signal to use a broadband fixed wireless service. But if the signal can bounce from house to house through my neighborhood, there is likely to be an unobstructed path that works. The redundant, shorter-distance links also reduce power requirements (important in mobile or industrial environments) and improve reliability. And since the mesh is implemented through software, it can operate across different kinds of hardware in heterogeneous networks.

For all these reasons, wireless mesh networking has been the subject of extensive research for many years. One such project at the University of California at Santa Cruz, in conjunction with SRI (a prominent California-based R&D organization), spawned no fewer than three mesh networking startups: Rooftop Networks, PacketHop, and SkyPilot. The US military has also been a significant supporter of mesh networks for their potential benefits to battlefield communications. DARPA, the military research unit responsible for the early development of the Internet, has funded several significant mesh networking projects over the past decade.

Until recently, though, the poster children for mesh networking had the bad habit of failing. Badly. Rooftop Networks was acquired by Nokia and quietly killed. Metricom's Ricochet, which had some mesh aspects, died an ugly death, was bought out of bankruptcy... and appears to be dying another ugly death. SkyPilot and PacketHop could only point to trials and grand plans for commercial deployment.

Wireless mesh networking seemed like one of those engineering concepts that looked great on paper, worked in the lab... but couldn't make the jump to the real world.

Meshes Get Real

Over the last year, mesh networks have finally started to click. A number of startups are deploying broadband meshed wireless systems, while others apply the mesh architecture to distributed sensor networks.

MeshNetworks, based in Florida, won't win any prizes for the originality of its name. But the company, which owns the commercial rights to technology originally funded by DARPA, is making good headway in the market. It announced a city-wide deployment last month for city workers and public safety officials in Medford, Oregon. Competitor Tropos Networks, funded by major investors such as Intel and Benchmark Capital, has rolled out mesh networks for police departments in San Mateo, CA and North Miami Beach, FL, as well as a citywide public deployment in Cerritos, CA. And PacketHop established a mobile mesh overlay network for state and local emergency services agencies in San Francisco, effectively creating a private wireless Internet for public safety activities around the Golden Gate Bridge.

As these examples show, public safety organizations such as police and fire departments have emerged as the primary early market for mesh wireless systems. These organizations need instant, secure mobile data connections, but today they are stuck with multiple incompatible legacy systems that don't even provide the necessary capacity. With homeland security a pressing concern throughout the US, public safety agencies are in a position spend money on emerging technologies that solve difficult problems. Corporate and residential deployments may take longer, but once the technology is proven, they can't be too far behind.

Another key reason mesh networks have finally taken off is the success of WiFi. With WiFi radios standardized and increasingly cheap, mesh networking companies need only wrap their own software around commodity hardware. Building scalable mesh networking software is a significant technical challenge, but today companies can focus on that challenge rather than having to build an entire proprietary system. At the lower end of the market, companies such as Locustwold in the UK are selling inexpensive WiFi access points with meshing software built in, making it easy to extend wireless networks beyond single-location hotspots.

Meanwhile, startups such as Ember and Dust Inc. are working on completely different applications for mesh networking. They are developing low-power ad-hoc wireless sensor networks. With a mesh architecture, a large number of inexpensive sensors can be deployed throughout an industrial or other environments. The devices find one another and establish a network, without any predetermined map or hardwired connections. The nodes can be embedded into industrial lighting or heating systems, connected to surveillance cameras, or spread across a battlefield. Both Ember and Dust trace their roots to university research programs -- Ember at MIT and Dust at the University of California at Berkeley - and both have raised substantial venture financing.

The Meshy Details

All this is particularly significant because of how well meshes, er, mesh, with certain kinds of applications. The first wave of mesh networks simply provide basic connectivity, but the mesh architecture will inevitably lend itself to certain uses.

In particular, mesh networks are perfect platforms for social networking applications. Social networks also involve distributed, ad hoc connections among users on heterogeneous devices. As meshing software becomes more widespread, we could see flash mobs on steroids, new location-based social applications, and tools to support virtual communities sharing photos, geotargeted annotations, and more. On the other hand, those same mesh networks will be used for pervasive video surveillance (by the government and by private actors) and for privacy-invading tracking of all kinds of activities.

Like it or not, mesh networks are here to stay. And as we saw with the Internet, new connectivity platforms can spawn all sorts of unexpected applications. The main thing holding back wireless mesh networks has been proof of their commercial viability. And that's the barrier that seems have fallen during the past year.