EyeForWireless: What A Mesh
This panel discussed the subject of Mesh networks. Formerly a military network technology used to rapidly deploy ad hoc, robust networks that could self-repair after the destruction of any particular node, mesh is now a way of deploying a network where each customer’s connection (a node) bounces from node to node towards a node that also serves as a backhaul point and delivers the IP traffic to the net backbone. The efficiency of Mesh networks depends largely on the efficiency of the routing algorithms in the nodes. How fast can each node receive, and pass along a packet?
A good argument in favor of mesh that made by Sri from Tropos was that in any RF technology, the link-budget (the received signal strength) is directly related to the throughput – a stronger signal means more throughput. Therefore, with a few centralized towers, you cannot deliver consistent, fast throughput to an entire Metropolitan area (the closer receivers could be faster, the further receivers slower). With a mesh, since there are many more nodes in a network, the receiver is (theoretically) always closer to the nearest fellow node, and therefore will receive a stronger signal than from a distant tower. I say “theoretically” because this makes the big assumption that a lot of mesh nodes exist. Mesh networks, unfortunately, suffer largely from the “chicken and egg” problem, in that they offer little value until a network us largely deployed. Yet deploying the network requires getting a LOT of nodes in the network, and getting real estate for nodes is problematic. Why would people become a node in a new mesh network?
All in all, the session and Q&A mood seemed to be one where Mesh is a network technology that’s real, but there are relatively few real mesh technology networks.
Comments on “EyeForWireless: What A Mesh”
overhyped
Mesh networks, based on WLAN technology, still face the fundamental problems of latency and throughput. First, you must realize these are half duplex connections; either receiving or transmitting at ony one point in time. This combined with the TX/RX latency of the radio kills throughput to less than 50% for each relay point. One often overlooked aspect of this is people believe that only the end node users see this degradation. This is a common misconception. The first relay node user has to carry the traffic of all other nodes throught them as well as their own data requirements.
As radio systems get faster (throughput data rates) you will be able to load more nodes and segments. Until then, you can realistically support very few users on a system like this. Take a look at the Nokia Rooftop user experiences.