Mesh Networks: The Counterarguments

While freely acknowledging the revolutionary potential of the mesh, you will see the following
challenges facing those who utilize mesh architectures in full-scale metro deployments:
Developing routing algorithms that maximize network efficiency and minimize congestion,
unacceptable latencies, and “router flaps” where highly inefficient routing
paths are inadvertently selected: Packet routers function imperfectly in public networks
today and do not provide the same kind of wholly predictable network performance as do
traditional circuit switches. Indeed, popular protocols for improving router predictability,
such as Multiprotocol Label Switching (MPLS), DiffServ, and Reservation Protocol (RSVP),
make the operation of the router notably more switchlike and set up fixed paths through
the Internet that essentially ignore the basic mesh structure formed by the crossconnections
among the major peering points.
Securing the network against hacks and malicious code: If the basic neurons of the network
are exposed, which they certainly are in any peer-to-peer arrangement, then it is
difficult to protect the core management software. Advocates of grid computing have been
preoccupied with network security, and rightly so, but really satisfactory solutions have
not been developed to date. Essentially closed systems such as class 5 circuit telephone
switches, ATM switches, and optical switches have generally been highly resistant to intrusion
and compromise whereas IP routers are almost routinely subject to successful
attacks. There are those who argue that nonhierarchical systems will be more resistant
to attacks, and the effects of such attacks will be localized rather than catastrophic, but an
absolutely convincing case has yet to be made for this position.
Minimizing hops: The longer the chain of wireless routers going back to the Internet
access point, the less bandwidth available to any given node for transmitting its own data,
and this constraint ultimately limits the size of a mesh network. If a mesh network grows
sufficiently large, nodes on the periphery will have to go through multiple intervening
nodes, which severely limits throughput for any one node. This deficiency is exacerbated
by the inability of most radios to transmit while they are receiving.
Developing standards: Single-vendor solutions are the norm in mesh networking today, a
state of affairs that, ironically, militates against the kind of pervasive deployments mesh
advocates envision. Currently little comprehensive standards work is under way in the
area of wireless meshes, and no major telecommunications infrastructure manufacturer is
directly involved in mesh projects.
Providing a strong first-mover advantage and a real ownership of the network for the
network operator initiating the mesh: If the intelligence is in the subscriber terminal,
there is really nothing to stop subscribers from forming loose cooperatives and dispensing with the middle man, so to speak, or else reducing the middle man to a role of simply
negotiating and maintaining a link with an Internet Point of Presence. Mesh equipment,
inasmuch as it is nonhierarchical and self-organizing, inherently favors ad hoc, cooperative
associations of nodes, and some of the vendors have acknowledged as much. It is
rather difficult to ascertain what the precise role of the network operator would be in a
truly self-organizing mesh other than serving as a sole conduit for a monopoly infrastructure
product, and I cannot see anyone fulfilling such a role indefinitely. It well may be the
case that a mesh of meshes on a national scale is infeasible and that local meshes simply
cannot coalesce into an all-encompassing canopy, but if they can and do, the function of
the service provider must change radically. Technology pundit George Gilder speculated
on the properties that a nationwide fiber-based mesh network might exhibit in his book
Life After Television (W. W. Norton & Company, 1994), and he assumed that the service
provider’s role would consist merely of owning and maintaining the actual fiber plant, but
in a wireless mesh there are no actual links to own. The service provider’s place in such a
scheme is obviously at issue.
A final challenge is the attitude of entrenched wireline incumbents in the event that wireless
meshes begin to win acceptance. One could foresee such incumbents seeking some sort of
legislative or regulatory relief to preserve the de facto monopoly they enjoy today. And in the
United States one could easily see them prevailing against the interests of the public.
All of this does not preclude a network operator soliciting a bid or a proposal from a wireless
mesh manufacturer. All of the claims made by such manufacturers may yet be realized,
and meshes may eventually become the norm rather than the exception. I think, however, that
meshes, because they are relatively untried, represent a higher risk option than more traditional
architectures.