We have been trying to get a mesh configuration running using HWMP that is more scalable than simple WDS bridging.

We are using 3.17 on an x86 platform with the standard wireless driver package.

We created a mesh interface using the default values. We configured wireless interface with WDS dynamic mesh. All nodes are set to AP-Bridge and 2.4GHz B only. WLAN1 and ETHER1 were put into the mesh. IP address was applied to the mesh interface.

All routers were spread about 40 feet away from each other and we turned down tx gain to 10dBm with unity gain antennas so that there was no excessive noise. Noise floor was about -96.

First thing we saw were crashes of nodes. Isolated it to our having adaptive noise immunity enabled and running up the CPU utilization to 100%. Normally with point to multipoint we can run ANI and stay well below 50% CPU utilization. So we turned off ANI. No more crashes. ANI seems to not work well with mobile mesh types of configurations.

Our first test was with 4 nodes and immediately were getting very strange results. With a pc client attached to the ether1 port of one of the nodes, we would attempt to ping (IP) across the links. Sometimes we would get a response, sometimes not. We adjusted PREQ LIFETIME down from 5 minutes to 30 seconds and were able to get most of the mesh running. But it was not too stable. MAC addresses were very confused. MAC connections using telnet or winbox were terminating on the wrong devices, etc.

One thing we saw was that if we created indirect links between two nodes by dropping power and putting behind a wall so that the two nodes could not see each other, but had to transit an intermediary, we could ping between the two end nodes only after pinging the intermediary. So it looks like ARP resolution needs to be refreshed manually for the mesh to work. Seems like this should have been part of the roles of HWMP.

We added two more nodes, for a total of 6 and were able to keep the network working, but not very reliably.

QUESTION: Has anyone successfully configured HWMP with dynamic WDS for a mobile mesh type of application?

So then we went back to a standard bridged dynamic (not dynamic mesh) WDS network. Same 6 nodes, same frequencies, except that we put it into a bridge running RSTP with a more frequent aging time of 1 minute, instead of 5 minutes. It was equally unstable.

Then we enabled the wireless-test driver using the same dynamic WDS bridge configuration. It was rock solid with low ping times, fast recovery. All nodes pingable and no crashes.

Note that the wireless-test driver does not support mesh.

So, based on the findings that we have, HWMP is not stable for mobile mesh applications. WDS dynamic bridge with RSTP is stable up to 6 nodes for mobile mesh. Turn off ANI.

Does anyone have a configuration that scales better than this?

Let me rephrase my posting:

Has anyone successfully implemented HWMP mesh using WDS dynamic mesh with more than 6 nodes?

How?

Thanks

Hallo !

Thank you for posting your findings.

I am also implementing a WDS mesh network for a mobile application ( the APs are moving all the time). I am having the same problem of network instability and unreliability.

Did you find the solution to your problem. Since your post is very old how did you solve your problem.
If you are still active on the forum we could discuss the solution.

Thanks!