A network loop is a cyclic forwarding path that repeatedly carries traffic without reaching its destination. At Layer 2, Ethernet frames have no TTL-like field, so flooded broadcast, unknown-unicast, and multicast frames can circulate indefinitely. A Layer 3 routing loop differs because routers decrement IPv4 TTL and drop packets when it reaches zero. section 7.1.6 section 14.1
Redundant switch connections create valuable alternate paths but also create Layer 2 loops when multiple paths exist between LAN nodes. A broadcast storm can overwhelm switches, consume network resources, and cause loss of service for connected hosts. section 14.1 section 14.6
Three-switch ring
PC1 sends a broadcast to SW1. SW1 floods copies toward SW2 and SW3; both switches flood the copies onward, producing clockwise and counterclockwise frame loops. The frames repeatedly traverse the ring and may also be delivered repeatedly to end hosts. section 14.1
Prevent or limit network loops with these practices:
- Use Spanning Tree Protocol to block redundant connections, leaving one active path while retaining blocked links as failover paths. section 14.2
- Understand STP’s root-bridge election, root-port selection, and designated-port selection, which use BPDU messages to create a loop-free topology. section 14.6
- Treat PortFast as an end-host feature only; enabling it on a switch connection can cause temporary bridging loops. section 14.5.1
- Expect original STP to converge slowly after failures: default timers can require up to 50 seconds before a blocked path forwards traffic. section 14.4.2
