September 2005

From the Editor: Mesh Networking Matters

Meshes will be the mechanism by which machine intelligence becomes like electricity: invisible and ubiquitous.

By Jason Pontin

The benefits of any truly transformative technology are at first exaggerated, but their long-term effects surprise everyone. At the moment, mesh networks are experiencing such misvaluation. Their promoters (and they are many) now describe them with hyperbolic enthusiasm; but in the end they will be the mechanism by which machine intelligence becomes like electricity -- that is, invisible and ubiquitous.

Mesh networks are not so very new: their conceptual lineage dates back to packet radio, a kind of digital data transmission used by amateur radio hackers in the 1970s. But investments in more reliable and intelligent networks made during the 1990s by the U.S. Department of Defense renewed interest in meshes; and within the last five years, academic institutions like MIT's Media Lab and startups like Aeria, BelAir Networks, Ember, MeshNetworks (now owned by Motorola), and Tropos Networks have rapidly advanced the technology. (Disclosure: Ember's chairman and acting chief executive, Bob Metcalfe, also serves on Technology Review's board.)

Meshies believe that mesh networks will overthrow traditional networking and communications and create entirely new kinds of distributed software. For the purposes of this column, mesh networks (sometimes called mobile ad hoc networks, or MANETs) are local-area networks whose nodes communicate directly with each other through wireless connections. It is the lack of a hub-and-spoke structure that distinguishes a mesh network. Meshes do not need designated routers: instead, nodes serve as routers for each other. Thus, data packets are forwarded from node to node in a process that network technologists term "hopping."

Before dismissing mesh networks as being of interest only to specialists, consider their advantages over existing hub-and-spoke networks. Mesh networks are self-healing: if any node fails, another will take its place. They are anonymous: nodes can come and go as they will. They are pervasive: a mobile node rarely encounters dead spots, because other nodes route around objects that hinder communication. Meshes are cheap, efficient, and simple.

But they are still in development. The chief technical challenge for meshes is the inherent unreliability of wireless links. Because the unreliability compounds with each hop, the size of meshes is now limited. A related problem with hopping is that, for now, moving nodes seldom establish new connections "seamlessly": when a network's topology changes, some transmission paths can be temporarily disrupted. Therefore, voice and video sit unhappily on meshes.

Meshes lack standards, too: low-bit-rate mesh networking has a standard called ZigBee that is supported by around 100 companies, including Motorola, Mitsubishi, Phillips, and Samsung, but high-bit-rate communications have no such standard (although the 802.11 committee of the Institute of Electrical and Electronics Engineers hopes to create one by next May).