To Wireless Mesh or Not for range extension

I have one RBM33G board with a single radio interface R11e-2HPnD card installed and two omni-directional antennas. This OpenWRT configuration has 7 VLANs with up to 50 devices running traffic through it. Applications are wireguard, netdata, Luci, etc... Its acting as an edge router and WAP.

I'm not only wanting to increase its range, but also join it to another build that is almost the same, minus a few edge router apps. This one will act as a bridged WAP.

I'm questioning whether to go Mesh between the two or just create separate wireless networks. Eventually its going to end up with the following architecture:

<-------------------------Street Block----------------------->
Building A Building B Building C
[AP/Router1] <---50ft---> [AP2] <-----50ft----> [AP3]
|
|
Internet

1.) The RBM33G boards can support up to 2 R11e-2HPnD WiFI radio cards. Using OpenWRT, would I have to install a second card in each AP and have them communicate with each other on those radios? I currently have 4 SSIDs (separate) networks configured on one radio card with separate policies including SQM regulations on each wlan0-X interface.

2.) I noticed that with OpenWRT you can scan and connect to another AP with a single radio. Is this the same as a Mesh? Using the scan and connect method, could I daisy chain the illustration above by having AP3 connect to AP2 then AP2 connect to AP1 allowing all APs to get out to internet?

I'm green to Mesh, thank you any direction veterans.

With only two devices, you don't need "mesh", it's just point-to-point. If you've got three stations, then mesh with mesh routing can help in case one link goes down by providing a middle-man path.

Wired is the best option. If you can't run wires, users report that the better power-line modems work reasonably well.

Yes, if you go wireless, the best option would be a separate radio for the backhaul, so that you don't consume bandwidth and time slots from the one you're using for client access.

With a point-to-point link, using a GRE tunnel may be a good option. (You'll need something more than WDS to manage multiple VLANs over the link.) A Layer 2 tunnel will provide the possibility that users can "roam" from AP to AP without requiring new DHCP and the subsequent drops due to a different IP address. Another option would be a VPN link, which will generally also add encryption, which GRE, by itself, does not provide.

1 Like

Technically speaking, depending on the wlan cards you're using, you don't need one radio for the uplink and another dedicated one for the downlink, but performance is significantly better if you do use a dedicated card for the uplink and downlink. Using this topology WDS/ 4addr should do, with GRE on top to employ dedicated tunnels per VLAN.

2 Likes

So mesh is only feasible when you have all APs within range of each other or at minimum a pair in a group. Each AP having two potential routes to internet gateway. Thank you @jeff and @slh.

Are there any benefits to mesh architecture if applied in a daisy chain manner, each AP within range of 1-2, 2-3, 3-4-internet?

Regarding links, GRE through Wireguard tunnels sounds like a fun project.

Going with dedicated radios, shooting to accommodate VoIP traffic. Are using directional antennas outside on walls facing each other overkill if each point range is less than 100 ft in a point to point chain, or mesh? Or would 7-10 dBm omnis be sufficient, Type N connectors? The cards/radios claim a 30dBm output, add the wiring loss and brick walls loss, unsure but sounds like interesting testing.

Or would a single far reaching, narrow beam, from building A into the linear direction of the rest of the buildings, last in chain with direct tunnel to first?

The only benefit is its self-organizing nature, not performance if you do pair them manually.

1 Like

:unamused:

I'm not saying it can't be done but it seems unlikely. Wifi phones are already a little flaky, by the time you retransmit through a couple hops the jitter and latency are going to be trashed. If this is your goal I'd recommend point to point specialized links like ubiquiti airmax links. Use a star topology with the wired site at the hub. These kinds of devices use time division multiplexing so latency and jitter are controlled. You can put an Omni on the wired site and aim directional links from the other sites.

Each building is gigabit Ethernet to a dmarc in their building, each dmarc having omnis and/or directional hardware. Its a slow approach, time to test. Shooting for an entirely OpenWRT P2P and/or Mesh, with encrypted tunneling. Thanks for sharing of minds, will share results.

Only one straight line can be drawn between two points --from Euclid's first postulate

1 Like

Ah, but wifi thrives on ignoring that postulate :joy:

Yes, agree two points don't make a mesh , unless it is your starting use case and you intend to add more.

2 Likes

each building has its own gigabit ISP connection? What then is the point of the wireless?

No Daniel, to a DMARC (demarcation point) with a 24 Gigabit PoE port Switch in a closet. Each building having a back-haul to building A which does have a Fiber ISP gateway. Hope that clears up my bad ASCII text art diagram.

So that backhaul doesn't exist yet and you are trying to set that up using wireless bridges? And you want building C to backhaul by bridging to B and then B bridges to A. Is that about right?

What you want at those bridges is a transparent bridge hardware: ubiquiti airmax or the like. If you insist on 802.11 then you will have issues that it's not a transparent layer 2 bridge. In particular it doesn't transmit the MAC of the source it transmits the MAC of the device itself. The WDS method could help, except I've found it flaky and subject to just stop working randomly. 802.11s Mesh and BATMAN protocol handles this by doing routing computations at layer 2.

I think you are going to need BATMAN just to be able to compute the multihop path from building C back to A. I don't think 802.11s will handle that hybrid wireless/wired mesh stuff.

1 Like

Correct

Thank you, very informative. I do have plenty of time to test, so I'll share the experience. Trying out Building B to A in the next month or so. If all else fails, we're going to run Fiber joining each building. Sure would like to see an entirely OpenWRT implementation though. We do have two requirements for this, VPN being Wireguard and Monitoring being NetData nodes - its so nice to see OpenWrt supports them both now.

Here is some Sunday Eye Candy of the single Router/Wireless AP in Building A.

PS: Looks like I'm going to have to figure out how to free up some space on the AP to accommodate additional packages needed for WDS / BATMAN etc. I'm down to less than 2 meg of space.

In building A, the current AP/Router as mentioned above is in a closet on second floor about in the center of the building. I think what I may do today is place another AP in bridge mode in a room on second floor on wall facing Building B, strip out unneeded apps to free up space to prep reaching Building B.

[Building A]
[AP1} -- ethernet -- [AP2] <-- dedicated radio --> [Building B]

If I do this, it raises another question. If AP1 and AP2 are on the same Ethernet switch, whats the best way to handle the wireless on the AP2? I like the Mesh roaming abilities and wish to not have to have it beacon with different SSIDs. Is this point-2-point?

This is a common misconception. Mesh is a backhaul method. There are commercial "mesh system" products which also include AP functionality with (allegedly) improved roaming for the AP users. That is a separate issue though.

Unless the client devices could participate directly in the mesh they aren't "mesh roaming". Standard client devices like phones and tablets do not have any mesh drivers. They operate as clients of an AP.

Since you have a good wired inter-building infrastructure in place definitely use it. Fiber between the buildings instead of GbE would only be of advantage for very heavy LAN traffic such as having a NAS in one building being used by the others.

2 Likes

There is a server in Building A which has an SMB share on it (Windows Domain Controller :roll_eyes:) that will need to be accessed from building B.

On the mesh roaming note @mk24, so having two APs on the same Ethernet network is there any way to have both APs be transparent as one from a mobile device endpoint perspective? Single SSID. I misconceived the mesh notion this was part of its ability for phones, pdas, etc...

Or is it as simple as creating the same SSID's on each AP with the same passwords, but run each AP on a different channel?

Yes you can (and generally should) set all of your AP's to the same SSID and the same password, but try not to have physically nearby ones on the same channel.

Dropped into that environment, a client will generally at least start out by connecting to the AP with the strongest signal. One of the weaknesses of the WiFi standard is that the decision to roam was not specified and left up to the client driver, and most of them are not good at it.

1 Like

One reason to run fiber between buildings is that it doesn't conduct electricity. Honestly, I'd start looking at burying fiber between the buildings and using switches that support SFP modules, this is going to be so so much better.

I've considered Fiber, each switch in each building has 4 SFP ports, it makes sense. Though I've ran thousands of Ethernet lines and crimped RJ45 terminations, honestly wouldn't know where to start regarding running Fiber, much less how to terminate them. My understanding of terminating Fiber, is its a specialists job.

I think it makes sense to call around to networking shops and tell them you want them to run 2 fiber runs between adjacent buildings, and how much would it cost? The network will be way way way better with 2 fiber lines than with whatever you do with radio links. I imagine it'll cost you between say $200 and $400 total to get those two fiber lines run. By the time you're talking time and equipment and debugging and etc the radio links are going to be similar.

2 Likes