Maximum number of connected clients to a access point

#1

Hello All,

I have a little experience about the Networking. But need to know, how many devices connected to each other in a mesh-network(I think that each devices can talk to every other devices in network) over Wifi? Say we have a Jalapeno. It has dual band concurrent radio supporting 802.11ac Wave2. I think that Jalapeno support both 2.4 GHz (802.11b/g/n) and 5 GHz (802.11a/n). Then at most how many Jalapeno devices could I connect to each other over the their wifi? What are the limitation factors for maximum number connected devices? (hardware?, driver?, OS?) If Jalapeno supported a single band only (2.4 GHz (802.11b/g/n)), maximum number of clients would be also decreased?

Thank you so much,

#2

The limit you hit will likely be a practical one, not a technical one. Eventually the devices start "talking over" each other causing re-transmits and the throughput to drop faster than you'd like. At some point, the bandwidth and latency get to be too high for your application.

I'd guess it's in the dozens of devices somewhere for many applications. Probably a lot more if they aren't passing any data.

Is it really a "mesh"? That means that the devices all connect to each other, not having to go through some central point.

What are you trying to achieve? How much connectivity do you need? How much latency can you tolerate?

Edit:

Found

which suggests on the order of 40-50 per AP for 802.11n when the devices are using a single spatial stream. Another poster states "Rule of thumb is 25".

#3

I do not have a Latency tolerance, yet. And I do not have a min/ or max throughput per clients in each node. I have just three nodes, now and trying to reach to last node over the middle node. I may increase the number of nodes in network, as an mesh-network. Sure, it would not be a star topology. Just trying to know limitations about the node numbers and throughput. Is there a linear relation? i.e: Assume that 802.b/g/n provide a practically 90 Mbps bandwidth in a mesh network. Then I reserved nearly 3Mbps for each clients per node. Does it mean that I may connect up to 30 devices to each other?

#4

In all cases, my meshing has been Ad-Hoc, not Infrastructure...so there is no "Access Point."

The factors are: available bandwidth, distance optimization, RTS/CTS Threshold, noise, timing (e.g. beacon frames, etc.) for you and all other SSIDs on that single channel.

It has little to do with an individual piece of radio hardware.

As @jeff noted:

#5

I think you'll find that:

  • There is some "background" overhead, that you typically don't "feel" (unless you've got several SSIDs active on each radio)
  • It is reasonably linear with consumed bandwidth up to a certain point
  • As a secondary effect, is reasonably linear with number of clients up to a certain point (overhead of a connected client)
  • At some point adding more clients and/or consuming more bandwidth starts degrade overall performance (both aggregate, as well as per-client)
#6

In fact, I need to know whether Jalapeno has a maximum number clients limit. I think now hardware is no problem, but application specs (latency tolerance, throughput per node) are critical for max. number clients.

Sure it is mesh network. Why not? It is too costly? It should have been a star - topology?

Best Regards,

#7

Some people misunderstand "mesh" and think that it solves problems that it doesn't address, such as roaming of consumer devices.

It sounds like your clients will all be direct, mesh participants. If so, it can be a very good topology.

#8

What are peoples doing on wifi. Maximum is 3 if 2 are using unlimited torrent) Maximum is 15 if they on YouTube. 25 works well if all of them reading openwrt forum :slight_smile:

True is between 15 and 30 depended on clients hardware.

#9

Be careful with that, as even enabling 802.11b or "legacy rates" will drag down your whole network, even if all your clients support 802.11n. 802.11b is only good for 11 Mbps modulation, so 80% or so of that for throughput. 802.11g is only good to 54 Mbps modulation rates, again before real-world effects and overhead.

#10

The BATMAN-adv meshing protocol is designed with the potential for citywide meshes with thousands of nodes. But in that situation the nodes are rather "sparse"-- each node is only in radio range of a few neighbors. With a large number of nodes in the same room, it would reach a point where they don't all have time on a clear channel to transmit their beacons, let alone any traffic. I am not sure where that point is but it could be calculated.

The large number of users problem is really difficult to test. Perhaps just have a festival and watch what happens.

802.11a, the lowest standard on 5 GHz, starts at 6 Mbps. In other words, the b modulation and rates are never used on 5 GHz. They should never be used on 2 GHz either but don't count on being able to disable them.

#11

Jeff,

I think that not surely a mesh. Not every node on the network has to be connected directly. It should be like an ad-hoc networks? In which communication layer would Ad -hoc network be implemented? Could you explain the this point? I wonder that OpenWrt supports ad -hoc network architecture or not? Could Ad-hoc network be implemented on the only Software level ? Isn't there any specific hardware requirement for ad-hoc network?

Best Regards,

#12

Ad-hoc means that a station with something to send transmits the packet directly to the receiver. So this means that all the stations must be in direct radio range of each other. Think of an Ethernet hub (not a switch) operating on the air instead of with cables. Every packet that goes out is heard by everyone else, hopefully including the intended receiver. This was the first wifi mode, intended to be very simple and familiar and require no protocol software.

There is no provision for an intermediate station to relay the packet, which is what mesh does.

There is no control of who has permission to send at any given time. It works very poorly with a large number of stations. Think of a meeting with no chairman. Ad-hoc mode is almost never used in any modern network.

#13

The part of "is this really a mesh" that is of primary importance is the difference between:

  • Some type of connectivity interconnects a number of APs, clients then connect to these APs
  • Clients directly participate in the mesh, relaying packets directly from client to client

What is required in the first situation is "only" that the APs understand the mesh protocols and mesh routing. In the second situation, every one of the clients requires "mesh software and configuration". Most smartphones, laptops, and IoT devices do not natively support mesh membership and interoperation at this time.

Node-to-node connections such as "ad hoc" or 802.11s are Layer 2. The protocols that run over than transport, such as TCP/IP and route discovery are primarily considered Layer 3.

A "mesh" can run over various Layer 2 protocols, "ad hoc", 802.11s (with routing disabled), wired Ethernet, what have you. To some extent, the choice of mesh protocols (such as topology discovery and routing) is independent of the choice of Layer 2 connectivity (802.11s routing being a notable exception).

#14

mk24,

Ad-hoc means this really? Each station (node) need to directly see every other station in network? I think that ad-hoc means that a node could be reached to a non-directly connected node via a node where between of them. So, it is not necessary every node to be 'directly' connected in network. In ad-hoc, each node should be a little more autonomous. A node could be transmit, receive or route incoming message to correct node. And ad-hoc network be able to dynamically change itself when even a few node lost. I think your suggested Batman is a kind of ad-hoc network ? So this means that OpenWrt does not support the ad-hoc mechanism? We need such a Batman?

Best Regards,

#15

"ad-hoc" is purely a point-to-point connection. Routing and all related to it need to ride on top of the links. Yes, you'll need some type of routing process for any TCP/IP connections that aren't directly connected over a single link, be it B.A.T.M.A.N, OLSR, or the like.

Edit: To clarify a bit more, "ad-hoc" in the context of 802.11 is a common term for IBSS connectivity. See, for example, https://en.wikipedia.org/wiki/Service_set_(802.11_network)#Basic_service_sets_(BSSs).

It can also refer to "unmanaged" or "self-managed" networks, where there are, among other things, peering authentication, topology discovery, and routing services that allow the network to self-form and adjust its behavior as nodes come and go and link quality, connectivity, and load change.

#16

802.11 "ad-hoc mode" does mean this. The routing of packets over multiple hops is what the BATMAN etc offers running on top of "ad-hoc mode"

#17

Dear All,

So, below is given a hierarchical view;

Wifi Hardware <--> Ad-Hoc mode <--> BATMAN <--> OpenWrt

Is this correct, right?

Best Regards,

#18

It could be useful to look at the OSI network conceptual model https://en.m.wikipedia.org/wiki/OSI_model

Wifi hardware provides layer 1, ad hoc mode (IBSS) is one type of layer 2 link, Managed mode (access points) are another layer 2 method.
IP with static routes is a type of 3rd layer protocol, BATMAN over IBSS provides a different type of layer 3, as does OLSR.