Daisy chaining APs or connect them into the central router?

Question

I have a question about connecting APs to the main router.

In the picture below there are two options. Both options are connected entirely by ethernet cable and the access points act as ethernet switches:

1. In Option A, the modem (ISP), main router, and three APs are daisy-chained together.

2. In Option B, the APs are each directly connected to the main router which is in turn connected to the ISP modem.

Which is the better option for my home? Currently, it is not possible to wire each AP back to the main router; what are the risks of daisy-chaining?

Answer 1

Each "router in AP mode" is functionally like an Ethernet switch with an AP dangling off it. So you'd essentially be chaining those Ethernet switches.

Assuming each router has 1Gbit Ethernet, I don't think it should cause any significant problems as far as network performance goes – each switch should be capable of hardware forwarding between the "LAN" ports (unless it's really cheap); the router's CPU doesn't get involved and the latency impact should be very minimal as well.

Though it does limit you a bit with regards to where you can place the APs; some additional Ethernet switches (i.e. chain through a dedicated switch, with the AP as a "spur") might be needed to find a good balance between unobtrusive cabling and unobstructed APs.

And keep in mind that disconnecting one AP in the middle will make you lose network to the rest as well – they don't "pass through" when powered off.

Also, document literally everything as far as cables go. If an Ethernet cable goes from room A into room C, write that down on each end of the cable, and make good use of that network diagram tool you have.¹

So try to make direct connections to a central switch whenever possible, daisy-chain when that's the only option, and the larger chunk of your network needs to be chained off some device, the more reliable that device should be (e.g. sometimes chaining through a 1Gbit Ethernet switch is a better idea.)

---

¹ This recommendation brought to you by the utter mess that our predecessors had left us at our workplace, with no documentation or anything. "Why is there no network in classroom 307?" [an hour later] "Oh, it's because there is a 100M switch under the sofa in the teachers' lounge three rooms away and someone unplugged it to plug in a kettle."

Answer 2

It might depend on a few things.

If you're daisy chaining routers in AP mode - don't forget any traffic through AP1 needs to pass through AP2, and any traffic through AP3 needs to pass through the other ones. The wiring is going to be neater but you're theoretically reducing the maximum throughput of the routers down the line and should a router go down, anything down the line from that is off. With lower than what the line supports it should be fine with a little extra latency, but if your routers are running at say a gigabit of traffic total, you'd have issues.

While its more 'cable' - the star configuration is a much more sensible choice than cascading between APs/Switches. Each ap can in theory move traffic at line speed to the main router, and if one device goes down, the ones down the line don't all go down.

Answer 3

I would recommend going with Option B for several reasons:

Fault Tolerance: In Option B, you have a more fault-tolerant setup. If one of the APs or the cables connecting them to the main router fails in Option A, it could potentially disrupt the entire network. In Option B, each AP is directly connected to the main router, reducing the risk of a single point of failure affecting the entire network.

Bandwidth and Performance: Option B allows for better utilization of available bandwidth. Daisy-chaining multiple devices in Option A can lead to network congestion and reduced performance, especially if multiple devices are simultaneously using the network. In Option B, each AP has a direct connection to the main router, potentially offering better performance and bandwidth distribution. In your concrete example (assuming 1Gbit) in A our wifi is maxed out with 1Gbit/s in Option B it can theoretically reach up to 3 times the load (assuming users/devices are evenly distributed)

Simplicity: Option B is a simpler and more straightforward configuration. It's easier to troubleshoot and manage because you have a clear understanding of the network topology.

While Option B is the better choice given your constraints, it's important to ensure that the main router has enough Ethernet ports to accommodate all your APs. If you run out of ports, you can consider using an Ethernet switch to expand the number of available ports on the main router. (when using a switch keep in mind similar bandwith bottelnecks as with chaining)

Answer 4

I would consider checking that the AP's are supporting mesh, to have better future possibilities. When daisy chained and only configured as AP (not repeater), if one goes down all after goes down in that current configuration.

Not sure if mesh AP's automatically change to WiFi if ethernet backhaul goes down, but I believe it at least should be easier to heal the network in that case.

Answer 5

Option A (daisy-chaining) can work just fine.

But

It's not really the best way to go about things, and Option B ("star" designs, with each device connecting back to a common core) should be preferred when possible.

I run the network for a small campus, and in our network I try to limit to one additional link beyond the initial connection back to the core device. So if we think about these as switches instead of Routers/APs (since that's how we're using them here), CoreSwitch might link to SwitchA, which links to SwitchB, but I wouldn't want SwitchB to then link to SwitchC. I'd find a way to have either B or C also go directly back to the core.

I do have a two or three cases when this still does happen, and again: it does work. But don't have this be the default way you do things, or the first tactic you look at when connecting new devices. In fact, I'm in the middle of a project right now to get one "SwitchB" device connected directly to our core instead of another "SwitchA" (via fiber, yay!), so the downstream "SwitchC", which was added recently to support an Esports room, will no longer have so many hops.

This is especially true if you have any wireless trunk links. Mesh networks can work okay, but it's putting a lot of extra traffic on the least-reliable, slowest, and most-congested medium available to you. I see mesh as a backup option only, for when the existing wired link is broken.

On more advanced networks, you go even further and could link the Core to both A and B, and one of those might link to a switch C, which in turn might link back to one of others, creating a loop. And if we stopped there you'd have a broken network, but then you take advantage of Spanning Tree Protocol and now you have redundancy, where the network can use the surplus line as a sort of hot spare, so if one of the lines is damaged the extra link can automatically step up to provide service.

Answer 6

For WiFi Option B is the right choice since every AP decreases the bandwidth by half.

This means that in Option A, users of the first AP will get half of the bandwidth of the main router, users of the second one will get one quarter and users of the third will get one-eighth.

In Option B, every user will get the same one-half of the bandwidth, so it is much more preferable.

Note that if all your routers were mesh routers, both options would have worked the same and all users would have received the full bandwidth (within reason of the total network activity).

---

For ethernet connection, there is no such penalty, but in Option A, adding additional hops between nodes will mean greater latency for communications, which will be greater for the second AP and more for the third one.

Depending of the quality of the APs and on the kind of network activity, this latency may or may not be noticeable in everyday usage. It can be of the order of 5-125 microseconds per AP, so for the third AP it will be less than half a millisecond for slow APs and much less for fast ones. You can conduct a network latency test for your network, if necessary.

Option B will minimize the number of hops and optimize latency and should be preferred fully or partially. It also minimizes the number of points-of-failure in the network.