On Fri, Jun 12, 2015 at 04:46:11PM +0000, Francois Gervais via Devel wrote:
> Hi,
> 
> We have been running a mesh network with 15 nodes for a while now and
> it seems stable. We tried upgrading to a 28 nodes mesh and it work for
> 20 minutes and then crashed. 
> 
> The mesh still look fine but there seem to be a storm of ARP who-has
> broadcast looping around. 
> 
> Is the concept of broadcast in a mesh network explained in the IEEE
> 802.11-2012 spec? I'd like to learn more about it to help us diagnose
> the problem.

See 9.32.5 "Addressing and forwarding of group addressed mesh data
frames."

The standard doesn't prevent or prescribe any particular method:

"Procedures that enhance the reliability or efficiency of group
addressed transmissions are outside the scope of this standard."

With the implementation in Linux, broadcast is done with simple flooding.
There is a cache (rmc) so that a given node will only send a particular mcast
once, but all nodes will see and forward the frame.  It would be nice to
have something like multicast-to-unicast conversion but that doesn't
exist today.

One suggestion: if you're operating the network in 2.4 GHz be sure to
set the mcast rate to something higher than 1 mbps.  Also, if you can
identify leaf nodes, you might consider making them non-forwarding to reduce
the traffic.

-- 
Bob Copeland %% http://bobcopeland.com/
_______________________________________________
Devel mailing list
[email protected]
http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel

Reply via email to