Mesh Advantages

The arguments advanced by current manufacturers in support of the mesh topology
are several.
First, total cost of infrastructure is purportedly far less per a given coverage area. Mesh
advocates claim that the cost premium associated with a routing capability in the subscriber
terminal dwindles to insignificance with mass production and claim that the elimination of
base stations with their right-of-way costs and their costly, limited production hardware, as
well as their redundancy and ruggedization requirements, will save the network operator a
substantial portion of his total equipment budget. Add to this the huge reduction in backhaul
costs, and the supposedly less-critical siting requirements for individual nodes, and the cost argument seems quite persuasive. One vendor, California-based Tropos Networks, which does
not support 802.16, quotes a $20,000 to $50,000 price per square mile for the total number of
terminals necessary to provide pervasive coverage.
Second, coverage is supposed to exceed that afforded by point-to-multipoint networks,
even those equipped with smart antennas. This claim appears plausible because of the routing
diversity inherent in a mesh architecture, but the degree of routing diversity naturally varies
according to the number and distribution of nodes in the network. A network with only a handful
of nodes does not fully realize the advantages of the architecture. Metcalfe’s law, which says
that a network’s value increases by the square of the number of nodes, is certainly true of a
wireless mesh.
Third, the siting of the subscriber terminal is significantly less critical than is the case with
a point-to-multipoint network because if the node cannot establish a good link with one adjacent
node, it can probably do so with another. Some mesh advocates go so far as to claim that
the network operator can largely dispense with site surveys and that self-installation on the
part of the subscriber can become the norm. Again, I see merit to this argument, but I point out
that here also the strength of the argument increases with the density of network nodes. A
mesh with only a few nodes simply cannot have a lot of diverse routing paths for each of them.
Some of the more adventurous companies promoting mesh networks are New Zealand’s
IndraNet, Massachusetts-based Ember Corporation, and Florida-based MeshNetworks (now
part of Motorola). They claim that meshes will bring about fundamentally new service models
where wireless connectivity will become completely ubiquitous. The mesh itself will be used to
support massively parallel grid computing where many nodes in the network will participate
not only in routing decisions but in analyzing other networks such as transportation systems,
the power grid, and security systems, to name just a few. Such a fully realized mesh network
will also exhibit advanced storage capabilities and will serve to host a tremendous multitude of
applications as well. IndraNet principals perhaps go furthest in touting the transformational
capabilities of wireless mesh networks, suggesting that they will bring cheap pervasive voice,
data, and video connectivity to every corner of the world and will exert revolutionary effects on
everyday life almost everywhere.
It is difficult not to react sympathetically to the idealism animating such speculations
(because who would not want the benefits of advanced communications made available to all
people?), but network operators striving simply to establish a profitable business in a given
market are still forced to ask how well a mesh approach will serve the present needs of establishing
an initial presence in a market and ultimately reaching the key customers necessary to
sustain the business.
Absent any significant number of successes in actual deployments, the answer to that
question is elusive, but I offer the following observations.
Distributed intelligence and parallel computing are no longer radical ideas. Major information
technology companies, including Sun Microsystems and IBM, are firmly behind the
notion of grid computing, and indeed the fastest computers made today are no longer expensive,
one-off supercomputers but instead vast assemblages of conventional CPUs linked by
software. That a computing grid could manage a metropolitan network is by no means inconceivable,
and that it could manage it more cheaply and effectively than a central office rack of
big iron switches and servers is possible though not proven.
Mesh deployments are said to have been used successfully by the U.S. armed forces in the
field, but the federal government has divulged little information on the scope of such deployments
and the traffic densities occurring within them. Nokia sold its Rooftop Communications infrastructure equipment to a handful of service providers, providing some commercial field
experience, but the Finnish company has since ceased production of the radios. Finally,
several service providers have set up wholly proprietary mesh architectures of which the nowdefunct
Metricom Ricochet network with more than 30,000 subscribers was by far the largest.
Whether these constitute real proof of concept is debatable because none of the service providers
has survived.