Just curious and was theory crafting a little while back and was trying to figure out AREDN's capacity. I know there are plenty mesh network bubbles and some tunneled to one another, but in a perfect situation where every node was RF how many nodes could AREDN support? How many hops away from node to node could someone reach? Could one theoretically have nodes end to end that could reach from one coast of the USA to to another like hopping on packet radio? How about a large state? A satellite (yes there would be Doppler but what about geosynchronous?) yes outlandish but curious...
Again I am not suggesting anything, and I know there are multiple factors in this question but I was just trying to wrap my mind scenarios.
Again I am not suggesting anything, and I know there are multiple factors in this question but I was just trying to wrap my mind scenarios.
We found that when networks grew, the older devices with slower CPUs and lesser amounts of RAM couldn't cope with the increasingly large tables that OLSR was propagating. Newer hardware pushes those limits out of course.
The other issue is that OLSR v1 is susceptible to route flapping (flapping back and forth between two almost equal-quailty routes to the destination). As networks get more tunnels and thus more nodes it becomes increasingly difficult to avoid those big long routing loops (think West Coast to East Coast to Europe to Australia to West Coast).
In Southern California we've tried to mandate tunnels for local use only, and discourage tunnels that lead to other tunnels (think hub and spoke topology rather than mesh).
A number of months ago our node count approached 700 and the network was down more often than it was up. Remediation was a mixture of tracking down a couple of nodes with issues, and eliminating some long-haul tunnels. That brought the node count down to about 400 and at that level it's fairly stable.
I know your question was primarily about RF links, but they're treated about the same in the routing software.
Hope this helps.
Orv W6BI