How A Mesh Network Decides Which Node You Belong To

A mesh system presents one network name across a house, and devices appear to move between nodes seamlessly. The handover is less coordinated than it looks.

The client chooses, not the network

In Wi-Fi, the decision to associate with a particular access point belongs to the client device. The network can suggest, but it cannot compel.

Devices evaluate signal strength and decide when a different node is worth switching to. The thresholds are set by the manufacturer and vary widely.

Many devices are conservative, because switching costs a brief interruption. A phone will often stay attached to a weak node while a stronger one sits in the next room.

Sticky clients degrade everyone's throughput

A device on a weak signal falls back to slower modulation, which means each of its packets occupies the airwaves for longer.

Because Wi-Fi is shared, that airtime is taken from every other device on the same channel. One distant client can noticeably slow a whole network.

Mesh systems therefore try to encourage movement rather than waiting. The methods available to them are indirect.

Nodes nudge devices towards better options

The simplest approach is to stop responding to a client whose signal has fallen below a threshold, forcing it to look elsewhere. It is blunt and occasionally drops connections outright.

More cooperative methods let a node send a message listing better candidates and politely request that the client move. Devices that implement it transition quickly and without dropping.

Support is inconsistent, particularly among older devices and appliances, so a mesh network typically uses several techniques at once and hopes one applies.

The backhaul is the hidden constraint

Each node must carry its clients' traffic back to the router. If it does so over the same radios serving those clients, capacity is consumed twice.

Chaining a node off another node compounds the effect, which is why a third hop often performs far worse than expected.

Systems with a dedicated radio for backhaul, or with wired connections between nodes, avoid this entirely. A wired node is nearly always the largest single improvement available.

Placement beats node count

Adding nodes to a poorly covered house often produces less improvement than moving the existing ones, because each additional node consumes airtime as well as providing it.

Nodes need enough signal between them to run a fast link, but sited too close they simply overlap and compete.

The useful test is not whether a node has coverage, but whether the link between it and its parent is fast enough to be worth the airtime it costs.