In simple Ethernet (without switches), hosts and other devices generate a large number of broadcast and multicast frames. These frames compete with data traffic for bandwidth, and the broadcast domain can become a bottleneck. Broadcast and multicast frames include those sent for address resolution, dynamic host configuration (DHCP), spanning tree protocol (STP), and Windows tasks.
To reduce the size of a broadcast domain, you can implement VLANs in a switched Layer 2 network. VLANs break up broadcast domains at the switch level and allow configuration by logical function instead of physical location. VLANs also provide security benefits if configured correctly.
With switch-based networking, each switch port is assigned a VLAN ID. Hosts connected to a particular switch port are members of that VLAN, and only frames destined for other hosts on the same VLAN are forwarded to the connected host. This reduces the amount of unnecessary data traffic, preventing the hosts from listening to and receiving unwanted broadcast frames.
VLANs do not eliminate collision domains, however. Each half-duplex segment between a host and the switch is a collision domain, and frames may still collide at one of these boundaries. As a result, you must continue to employ traffic management practices such as STP and DHCP helpers to maintain efficient network performance.
In addition to VLANs, you can improve performance by implementing switch-based layer 2 switching technologies. For example, Cut-Through switching uses a store-and-forward method to reduce the size of a broadcast domain by examining each frame for errors. In contrast, a switch using Fragment-Free switching examines only the first 64 bytes of each frame before forwarding it to the next device on the network.