full-duplex operation
Full-duplex means two parties can talk and listen at the same time, the way a phone call lets both people speak and hear at once. Its opposite, half-duplex, is like a walkie-talkie or a single-lane bridge: only one direction can be active at a time, so you take turns. The terms describe whether a link can carry traffic both ways simultaneously.
On a modern switched Ethernet link there are separate wire pairs (or separate fibers, or separate frequencies) for transmit and receive, so a machine can send and receive at the same instant. With no shared medium there is nothing to collide with, so full-duplex links do not use CSMA/CD at all — collision detection becomes unnecessary. A 1 Gbps full-duplex link effectively gives 1 Gbps in each direction at once. Half-duplex, by contrast, happens on a shared medium (a hub, old coax): both directions share the same channel, only one station may transmit at a time, and CSMA/CD with backoff is required to handle the inevitable collisions.
This is one of the biggest practical wins of switches over hubs. A device connected to a switch port negotiates full-duplex automatically (auto-negotiation), doubling effective capacity and removing collisions and backoff delay. You will still meet half-duplex on shared wireless: Wi-Fi radios cannot transmit and receive on the same channel at the same time, which is one reason Wi-Fi uses collision avoidance rather than the wired collision detection.
Two machines on a switch link both upload and download large files at once: a 1 Gbps full-duplex port carries roughly 1 Gbps up AND 1 Gbps down simultaneously. On an old half-duplex hub, the same pair would have to share one channel and back off whenever they collided.
Separate send/receive paths mean no collisions and double the throughput of half-duplex.
Full-duplex does not lower latency — it lets both directions be busy at once. And a duplex mismatch (one side full, one side half) is a classic fault that causes errors and terrible performance.