One switch, many networks
Without VLANs, every port on a switch is one segment: every device sees every broadcast and can reach every other device at layer two. A VLAN assigns each port a number, and the switch treats ports with different numbers as if they were on different switches with no cable between them. Between two switches, a single link carries frames from many VLANs by adding a tag to each frame that names its VLAN; that link is a trunk, and the ports facing devices, which carry one untagged VLAN each, are access ports. The tag is four bytes, the standard is old, and every managed switch supports it.
A VLAN is a layer-two boundary and almost always pairs with a layer-three one: each VLAN gets its own subnet, and a router or a layer-three switch moves traffic between them. That pairing is where the security value lives. Traffic between the user VLAN and the server VLAN has to pass a routed hop, and a routed hop is a place to filter and log; traffic within a VLAN passes only the switch and is invisible to anything else. Servers, users, guests, building systems, management interfaces of the network gear itself: each is a VLAN with a subnet, and the firewall rules between them exist only because the boundaries do.
The exams treat VLANs as core material. The vendor-neutral networking exam asks what a trunk carries and why a device in the wrong VLAN cannot reach its gateway; the Cisco associate paper has you configure access and trunk ports, name the native VLAN and troubleshoot a mismatch; the security papers ask why VLAN hopping is a risk when trunk ports are left at their defaults. The practical version of all three is that a VLAN is only a boundary if the ports at its edge are configured as such.
