One of the compelling cases FOR Meshtastic and the network of LoRa devices that has been organically built has been “in case the grid goes down.” Our fragile telecom infrastructure has failed us (at least regionally) as recently as this morning - look no further than the storms that ripped through Dallas this morning and this weekend. Ohio and Oklahoma I believe had some rippers too. Don’t forget the ATT outage earlier this year and then Verizon failed in Arizona shortly after that.
But the Mesh Network is very ah-hoc and unstructured. How would it handle the massive influx of traffic if we lost cell connectivity?
Proposing…
What if we all picked a time, could be 9:30 AM on a Saturday in your time zone.
Everyone send three messages.
Message 1- Public Channel
Message 2 - Private Encrypted Channel
Message 3 - Direct Message.
Will that impact the network’s ability t handle traffic?
What can we learn?
How can we objectively measure the event?
It’s already unreliable above 20% ch util, since nodes start to throttle.
2 Likes
In general I feel the ‘general chatter’ on the network probably stresses things almost as much of this. Lots of packets flying around.
As for measuring, you would need a way to capture traffic from at least a proportion of the nodes.
Just looking on one node wouldn’t be very clear, as wouldn’t know how many messages missed.
It might be possible if multiple nodes had MQTT ‘uplink’. Then someone could gather the data and see how far messages spread.
… in fact I guess it might be possible with a passive monitoring, just see how far the existing messages ‘spread’.
(I think a specially configured client that is watching the MQTT topic, could see where the same message is uplinked by multiple nodes, and do the calculations. It would have to be recording, rather than just looking at messages in the current server)
If didn’t have MQTT uplink enabled on multiple nodes. then could perhaps just capture the data via rangetest.csv files. Would need someone to collaborate on collecting a bunch of files, from nodes, to run the analysis.
Would be an interesting project. Perhaps the hardest thing would be finding some candidate nodes. Would need mesh that normally has good connectivity, ie typically reliable comminication between nodes (that the hoplimit, does allow end-to-end communication
1 Like