Seems to be proprietary dual radio modulation scheme not 802.11
The idea is that both networks can send an receive, but you spread traffic between them randomly, to reduce the chances of collisions, and thus increase bandwidth and reduce latency. In a lab scenario it is probably not as optimized as the solution proposed by OP; in a real case, it is probably as good as it can be.
It's point-to-point, not many-to-many.
Also, notice that this will only work if nodes are close enough to fill the gap between two adjacent nodes (or they will not be able to communicate with each other), but far enough so two non-adjacent nodes are always out of range (or they will interfere with each other when transmitting).
Now that sounds really difficult to grantee!
Do not complicate life and pull cable everywhere!
+1.
Even though you could emulate full-duplex wireless using the methods as shown in the OP's first diagram, things start to quickly break down when you start talking about topologies with the number of routers greater than 2.
As @mbo2o pointed out, you could organize the mesh into a ring, but then in reality, you would need a dual ring, in an arrangement similar to a FDDI ring. In order to achieve such a dual ring, each router node needs to "dual-attach" to the ring, which means that you need 4 radios just to connect to the rings. Then, at least one more radio for your devices to connect to the router. That means that each router would need to have 5 radios in it, which is getting into the realm of the ridiculous.
So instead of attempting to create a "wireless FDDI ring" (which actually would make for a pretty interesting research topic if someone wanted to publish a paper in an academic journal), just ditch the wireless, and pull cable everywhere.
The articles mentions transmitting using sectors (directional antenna) to be more like cellular networks
I think you could accomplish this with OSPF routing and manually adjusted link weights. Make two "colors" of router, red has lower weight on channel A and black has lower weight on channel B, they will naturally route ABAB etc.
2 Likes
I am not aware of the term FDDI ring but I will research on it tomorrow.
For now as @dlakelan correctly proposed routing with OSPF would be the solution to emulated full duplex wireless links.
here is a mikrotik tutorial that if adjusted to OpenWRT would do the trick
https://wiki.mikrotik.com/wiki/OSPF_to_simulate_full_duplex_links_with_redundancy
OSPF does the internal routing decisions and batman-adv the mesh routing
At Freifunk we prefer to combine cable and wifi mesh on both bands with a penalty system and batman to use only the link with the best quality. It look like this at my local community: https://karte.freifunk-emscherland.de/map/#!v:gi Check the gluon firmware: https://gluon.readthedocs.io/en/v2018.2.x/
This setup is more about range.
1 Like
Any updates on this?
I would not say the discussion is still open... do you have any specific question?
Wanted to know how to implement this along with batman adv.
You might want to create a thread to inquire, then.
If they do it, it is possible.
About using radios for rx or tx only, it's a way to mitigate the collision problem. 802.11 uses CSMA/CA (CA := collision avoiding); ie: if 1 radio only uses the channel, no collision may occur. But their protocol suffers the 'hidden node' problem; it may be the reason half throughput is gone at 10 hops distance. This is possible with default openwrt distribution and the routing deamon for that protocol.
About streams, you are probably referring to MIMO radios; ie: 'spatial streams' from 802.11N to 802.11AX. Streams management isn't available in userland; it's something the radio manages by itself. By wifi alliance standard. You can, of course, go out of standard. In order to have this working on openwrt, you need the routing daemon and some kind of custom driver for the radio (ie: it's not a wifi radio any more, it's another protocol using the same frequencies).
But there aren't silver bullets in radio mesh networking; ex: some other users wrote about 'wasting the 5ghz'. I'd say instead, that 3 radios aren't enough for trunking and distribution.
For a complete coverage of a bidimentional surface, you need triangles/exagons (ie: 3 channels / 6 channels) for trunking; 1 or more channels for distribution to clients, depending on terminal density (ie: if you have 3 users per square meter or 100 per square meter... in a tall building). This is what mobile companies do to serve our phones...