Types of Meshes

Mesh architectures take two basic forms: the switched mesh and the routed mesh. The second
predominates today.
In a switched mesh a more or less fixed pathway through the network is established for
each node, and each packet will follow the same path through the course of a transmission. In
the event of a severe degradation in the link, the subscriber terminal will renegotiate routing
arrangements with its fellows and establish a new pathway through the network, but this function
will be more in the nature of a restoration mechanism than a means of managing traffic.
This configuration is also known as an infrastructure mesh.
A routed mesh, on the other hand, will have active, intelligent nodes, which continually
evaluate link conditions and establish new paths on the packet level as warranted. As with the
core routers used in the public Internet, mesh nodes will weigh link performance across a
number of parameters, including throughput speed, traffic density, latency, packet loss, jitter, and noise, and will make determinations based on the requirements of the individual transmission.
Another term for this is the ad hoc mesh.
Routed mesh networks take a number of forms that tend to represent points along continua
rather than rigid distinctions. At one extreme, the all-knowing mesh is where every node
sees every other node and maintains complete routing tables. At the other extreme lie those
meshes where each node knows only its immediate neighbors. Intermediate forms are possible
as well where a single node may know a segment of the network extending beyond adjacent
nodes but not the entire universe of nodes.
All-knowing meshes, by providing each node with maximum information on the condition
of the network, enable each node to plot routes that best meet the requirements of the
transmission. On the other hand, the memory and processing power required to store routing
tables for the entire network and run global routing algorithms drives up the cost and complexity
of each subscriber terminal. And because each node must poll every other node, the network
overhead will be higher as well, increasingly so as the number of nodes increases.
Another pair of distinctions concerns the overall behavior of the nodes in performing
route computations. Some meshes are entirely reactive, only polling their neighbors when a
packet is to be sent. Others are proactive, looking at the network even when they are not transmitting
data payloads. Here again the distinctions are not hard and fast, and various degrees of
reactivity and proactivity are possible within the same network. Obviously, the more network
diagnosis the node is doing, the less bandwidth is available for data.
The routing algorithms utilized by the surviving equipment manufacturers are all
proprietary and extensively optimized for the wireless metropolitan area network (MAN) environment.
They tend to diverge sharply from the routing protocols used in wireline core and
edge routers. The maturity of wireless mesh routing schemes is a matter of considerable
debate, and the lack of mass deployments forces the network administrator to depend on computer
simulations for determining the appropriateness of any given mesh approach for a given
market. Unfortunately, according to the relatively few engineers with extensive experience in
designing and configuring meshes, simulations are often inadequate and misleading.
Routed mesh equipment is obviously more complicated to design and execute than is its
switched mesh counterpart, but its ability to adapt to changing network conditions is greater.
Today the majority of surviving mesh vendors are offering routed mesh solutions, perhaps
because initial wireless mesh research was done in support of ad hoc military deployments
where speed of deployment, noncritical placement, and overall network robustness were
the primary considerations and where the expense of using multiple routers was not an overriding
concern.