I have been hearing a lot of debate on this issue back and forth on packet collision. I am hearing that if you run multiple nodes as to relay to other stations, that you will have packet collisions if they are running on the same frequency. I will make this more visual. A<->B<->c, so A sees B, and b sees C, because B has lets say two separate rockets and sectors pointed in opposite directions. If you DTD the 2 separate radios at site B, does that not communicate between the 2 rockets at site B to make sure that they dont colide with each other?
Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer
I wouldn't get too hung up over your network not looking like a mesh. To begin with, you need to define a path for data to pass because you don't have the node density to do it any other way. In the end, the network may or may not pass traffic the way it was originally intended. But that makes it more robust in that there will be multiple paths to get to the same point.
I would say focus on the most efficient ways to obtain your initial objective(s) (perhaps defined by a served agency).
Many of us have had great success in approaching this entirely as a traditional network (backbone node, distribution nodes, deployed/user nodes). You might consider linking your mountain nodes on another band (3 GHz is what I use), and I also use 3 GHz from the backbone down to the distribution layer (I also call these mid-mile nodes) and then 2 GHz (channel -2) from there on down to the users. Other prefer to use 5 GHz where I've used 3 GHz.
I'm sure eventually node density will redefine this initial architecture an I'll be fine with that. At least, when all else fails, I'll have confidence traffic will get through the path I've originally defined.
Andre
When 2 links are sharing the wireless channel, and these links are part of a tower site where traffic is hopping through, then the throughput is 50% if perfect conditions as compared to using different channels. Both links can't transmit at the same time when sharing the channel and are taking turns. The 802.11 protocol (adhoc mode in use by AREDN) uses CTS/RTS method to avoid collisions (using CA = collision avoidance, not CD = collisions detection) and to mitigate the "hidden transmitter" problem that comes in to play here. I have been monitoring this protocol exchange recently on my own network with packet captures and inspection with wireshark.
Here is where the plot thickens and the story gets more interesting... I am not seeing this hidden transmitter issue be a problem with the A<->B<->C setup where A and C are hidden to each other and B is the tower site. CTS/RTS works to coordinate the traffic.
What I suspect I am seeing and has been a problem is the "exposed terminal" problem. https://en.wikipedia.org/wiki/Exposed_node_problem . The good news is that ch -2 has opened up a lot of links that are not possible with the noise in ch 1+. The bad news is all the patchwork of links being created all over the geographical area is causing problematic patterns and throughput issues to occur. I've had occurrences where turning on a mesh node in one location, with no obvious dependencies, wiped out a major backbone link and traffic came to a standstill in another area. I used different channels to resolve.
If you are on an area mesh network and everyone is only using ch -2, then you are likely facing the problem that throughput doesn't scale and voip/video isn't working very effectively. If every node is seeing every other node, then there simply won't be a way to scale up traffic. It is an ideal 'mesh' from a tcpip routing option and redundancy perspective, but it will be an inefficient 802.11 layer design to achieve high data throughput. This is because the mesh routing technology in use (OLSRv1) does not have sufficient information of the 802.11 layer to make optimal routing decisions to maximize throughput. (We will all be very happy when there is improved technology to upgrade/replace :) ).
To scale up data throughput, look at channel isolation (3Ghz and 5Ghz gives a lot more channel options) with localized cell coverage areas (to avoid channel contention elsewhere) and channel isolation hopping through mid-mile, mountain, or hub relay sites.
Joe AE6XE