search

maconard / cps-iot_leader-election

Group project studying Leader Election, continued from the CPS-IoT course at Michigan Tech in Spring 2020.
0 stars 0 forks source link

Network Topo #2

Closed DavidWHallberg closed 4 years ago

DavidWHallberg commented 4 years ago

Look into Network Topo generation in the FIT IoT-Lab

DavidWHallberg commented 4 years ago

Looking into a tool I found to use topos in FIT IoT Lab

https://github.com/koalo/iotlab_topologies

Looks like it even creates drawing on layout.

DavidWHallberg commented 4 years ago

Was a dead end. The tool was creating topology layout to then implement not to implement a given layout.

Looking back in RPL protocol on RIOT (not native)

DavidWHallberg commented 4 years ago

https://github.com/iot-lab/iot-lab/wiki/Using-Foren6-to-diagnose-in-realtime-your-6LoWPAN-experiment

Guide to make graph of the network topology using Foren6

Still looking to customize RPL but doesn't seem easy

DavidWHallberg commented 4 years ago

Nope RPL automatically creates its own topo (similar to a tree called Directed acyclic graph) and I can't find a way to easily change it (if you can)

Looking now into what it would take to create taps similar to what we were doing in simulations

maconard commented 4 years ago

Dang. Lots of dead ends here. Topology is important to this sort of work so I'm kind of shocked they don't have a better control system.

Appreciate all the work you're putting into this!

DavidWHallberg commented 4 years ago

Looks like the best option would be to skip neighbor discovery and set the IPs of its neighbors if we are going to do a set topology.

Would have to add code to determine it neighbors based on its own ip (maybe it works on real devices?) I would think we would also have some sort of flag on each node that is can communicate with its neighbors and then signal its ready before we start.

We would also need to have "fixed" nodes that we use and we would have to consider the physical layout of the devices when coming up with the set topology.

RPL does work by making a tree structure that looks to be dynamic so we can look at using that in different cases as well.

I will look into it a bit more before our next meeting but that is where I am at now.

aebne commented 4 years ago

To do for next week:

maconard commented 4 years ago

Update: no simple solution appears to be available for topology control, so we have switched to using an overlay topology method involving a master node that discovers the network and sends topology assignments to each node.