Calling all wifi wizards

OK so I was in a sales meeting today and had the sales guy lay this on me

When you have multiple clients connected to a WiFi Access Point like in the top right corner of the graphic then all of the clients that are connected to that AP are operating at the slowest client speed. ex. the one client that is at the edge of the network and is performing sub-optimally will slow down the other clients.

Thoughts?

I find it hard to believe that APs can’t multiplex their clients in order to avoid that, but I haven’t looked into it in detail.

I mean, the clients aren’t dependent on each other, so assuming there’s no significant resource contention within the AP, and you’re not being limited by the physical number of radios available to receive transmissions I don’t see any reason why one client would have to wait for another.

I think linking APs via WiFi (5.8Ghz or any other band) is a bad and flawed solution. Hook them up to 1G (PoE) ethernet each. In the diagram, client to client communication can be 4 hops of Wifi…and it’s getting more the more complex the mesh gets (stuck with ring topology?).

2 Likes

The vendor is a mesh-networking company for Industrial/Mining applications.

Well, if Ethernet wall jacks aren’t a thing, because you’re relaying signals through some mining shaft…or Ethernet wiring is just not possible or economically for that company. You need to take the bitter pill. I’ve not seen a stretched out ring topology Wifi in a business use case, but I know how they work in consumer land.

Problem with ring and bus topology is that the backbone/backhaul can get very bandwidth constrained. And 5.8Ghz is nice…but range is noticeably shorter than 2.4GHz too.

1 Like

I thought the 2.4ghz band had fewer frequencies to connect on, and the 5.8 had more?

So I would have thought switch the frequencies

Unless one needs penetration from each node?

(obvs better to Wire the nodes)

It’s not ring it is suppose dot be “least cost” meshing. So basically the ach node is in charge of figuring out whick next hop has the best signal and using that.

2.4 has fewer channels but far better penetrating and distance power.
a device with a weak signal will have a huge amount of packet loss/retries, that can tank the performance of an AP for every client connected.

you can use 5.8g directional antennas to get around the distance issue if you need too.

I am not a wifi guy, just wondering. and the client seems to be a mesh company, so questions were genuine.

I presume, the "mobile equipment* makes use of the longer distance of the 2.4 band?

And there must not be too many clients to connect to each node, so contention/competition is not an issue?

:person_shrugging:

It seems like this stuff doesn’t work how I might expect it to either, so I’m officially clueless.

:yay:

So concurrent connections and speeds depend fully on the type of AP being used.

APs following 802.11ax and with MU-MIMO (multi user multi input multi output) will connect to devices at different speeds based on signal strength and support standards.

So a IOT devices with only a 2.4ghz 802.11g connection wont slow down or impact a brand new wifi 6e device.

What will have a negative impact in this setup is the back haul channel. Whenever possible DONT USE MESH, because its reserving spectrum for inter AP communication. Sales people will always say “it wont slow things down they are dedicated channels” but those channels could have been used for client connections if not for the back haul requirements of Mesh.

I dont know anything about this company, but I would look at TP-Link’s business AP’s as they will be better suited and wont restrict things down to 2.4ghz.

1 Like

This is what I was looking for. I think some of the info in his sales pitch was a little date.

dug around a bit on the site and im not entirely sure I understand their value prop.

What is the use-case here and why would you “need” a node based system over a traditional wireless deployment?

You have mining machines that move around a mine and have WiFi/mesh routers onboard and can bridge some of the communications? Probably augmented by base stations placed around the site?

Basically what I am trying to figure out for the company I am working for.

Yeah that is their traditional bread and butter. I am not sure if it is really applicable to what we are doing.

Wifi doesnt operate at the slowest client speed per distance and channel. It operate at the slowest common standard and even then wifi 6 changed that

A client far away wont slow down the client closes. But the channel conditions itself will warrant slower communication to that specific client

Their value proposition seems to be highly available wifi. each node has a backup redundant backhaul path. Loss of one link will not damage the backhaul ability. If the mesh is setup properly while speeds and latency will be slightly variable the use of an entirely separate band is more reliable than an intermixed mesh

3 Likes

Industrial/Mining also raises the question of EM interference. Walkie talkies, machines, motors…there is a reason why we invented fiber cables before it was a bandwidth/latency argument. It’s not just microwave ovens in the employees kitchen.

QFT

But if the requirement is to keep it resilient and guarantee availability at the cost of latency/bandwidth and the deployment of cables isn’t feasible, there is an argument for it.

This did indeed used to be the case, but is not the case with newer wifi standards.

However, most industrial and IOT equipment still runs on 802.11n or g and has not been updated and never will be updated, and sadly even the newest released products still use old chipsets running on “n” much of the time. Its pitiful.
So my guess would be that the wifi the clients are all connecting with is likely slowing down to the slowest client for all those devices running on the old standard.

3 Likes

(a “client” is called a “station” in wifi speak).

Strictly untrue

But yes, slow clients suck, think of it this way…, there’s only a 1000ms in a second.

Let’s say you have 10MB of data you want to transfer to each client and you have 5 clients, let’s say that normally that would take 100ms each. Your access point is done in 500ms, and other 500ms is just idle/not doing much.

Let’s say 1/5 clients is slow, and 10MB takes 600ms instead of 100ms. Are you slowing down other clients… no they’ll be fine, everything still fits inside 1000ms.

Let’s say 1/5 clients is really slow, and 10MB takes 800ms instead of 100ms… all of a sudden you can’t fit 5x50MB, … so, you have to reduce demand for air time, and transfer less than 50MB/s, which I guess is some form of slow clients slowing down the fast ones maybe, if you’re looking at it from a fast client perspective.

With WiFi there’s also various overheads that you can’t get around, and you need to reserve air time for these. Also, there’s this RTS/CTS thing with wifi, where access point gets to coordinate which station gets to speak at any given time, which improves efficiency when there’s lots of stations (as opposed to each station opportunistically waiting for a quiet period trying to transmit, hoping it doesn’t collide/overlap with another station and then backing off).


I’d ask the sales guy how does their solution compare with a comparable bunch of Mikrotik access points, which are probably cheaper.


Ok, here’s the deal, my 2c. To get wifi running, you need to power the nodes. If you’re stretching cables for power, stretch Ethernet alongside too. Mikrotik in particular has various power over Ethernet solutions for extending Ethernet (gper for example) and poe powered switches and directional wifi antenas for cheap, and some of their stuff is in rugged boxes, if you want “industrial” feel for “mining” and things. You’ll need to have an MSP / contractor maintain either system, they can deal with configuring hardware remotely on short notice easily, and if something doesn’t like being reconfigured/needs a spare, well - MSP can figure out what you need so you get to minimize both downtime and effort - it’s their job. For longer distances you can do fiber, but that requires someone who’s not going to plug it in backwards and would keep it relatively clean on site, where as rj45 cat6 is easier for random on site dude to deal with and does the job

And if you really want wireless backhaul, nv2 between Mikrotik nodes actually isn’t that bad when configured right, but you can mix and match those point to point links, whatever.

Go for cheap gear, pay/hire a smart dude to make sure your network runs 24x7 and is easy to maintain.

I used to work in the mining industry and I see no need for this type of mesh. Here’s why:

  1. Each Mesh node needs power to operate, so you have to run cables anyway, might as well run 1 ethernet cable with POE.
  2. Depending on the mine site you may need to think of “explosion proof” containers or such, wifi doesn’t do well on the EXD testing and often will fail, it needs a dedicated solution which a mesh network probably wouldn’t solve.
  3. In my opinion it would be safer to run the lower voltage POE power for devices than the 110 or 240 to places where cables get abused.

Just my opinion, but in terms of the actual wifi meshing, yah can be done, but will stupid slow in a mine where obstacles are everywhere, so just use a cable.

1 Like