Redundancy is required
A single link or switch is a single point of failure. Real networks give every access switch two uplinks and connect distribution switches to each other, so traffic has somewhere to go when a cable is cut or a switch reboots.
…but Ethernet loops
Switches flood broadcasts and unknown unicasts, and Ethernet frames have no TTL. So redundant paths become loops:
- 1. One broadcast… PC A sends an ARP request.
- 2. …flooded both ways… SW2 sends a copy to each neighbour.
- 3. …and around, forever. The copies keep circling and multiplying: a broadcast storm, MAC tables flapping, and duplicate frames.
Two ways to make redundancy safe
| Spanning Tree Protocol | EtherChannel | |
|---|---|---|
| Idea | Block just enough ports to remove every loop | Bundle parallel links into one logical link |
| Backup links | Idle until a failure | All carry traffic |
| Works for | Any topology (rings, triangles, meshes) | Links between the same two devices |
| Learn more | Spanning Tree course | Next lesson |
- 1. With Spanning Tree: SW3 blocks one port, so every frame has exactly one path. If a link fails, the blocked port takes over.
In practice they work together: EtherChannel bundles the links between each pair of switches, and Spanning Tree still guards the overall topology, now seeing each bundle as a single link.
Check yourself
Predict · scenario 1
What lets a switching loop continue indefinitely?
Predict · scenario 2
Two switches are joined by two cables. Without EtherChannel, what does Spanning Tree do?