In early network implementations, it was common for organizations to have all computers and other networked devices connected to a single IP network. All devices in the organization were assigned an IP address with a matching network ID. This type of configuration is known as a flat network design. In a small network, with a limited number of devices, a flat network design is not problematic. However, as the network grows, this type of configuration can create major issues.
Consider how on an Ethernet LAN, devices use broadcasts to locate needed services and devices. Recall that a broadcast is sent to all hosts on an IP network. The Dynamic Host Configuration Protocol (DHCP) is an example of a network service that depends on broadcasts. Devices send broadcasts across the network to locate the DHCP server. On a large network, this could create a significant amount of traffic slowing network operations. Additionally, because a broadcast is addressed to all devices, all devices must accept and process the traffic, resulting in increased device processing requirements. If a device must process a significant amount of broadcasts, it could even slow device operations. For reasons such as these, larger networks must be segmented into smaller sub-networks, keeping them localized to smaller groups of devices and services.
The process of segmenting a network, by dividing it into multiple smaller network spaces, is called subnetting. These sub-networks are called subnets. Network administrators can group devices and services into subnets that are determined by geographic location (perhaps the 3rd floor of a building), by organizational unit (perhaps the sales department), by device type (printers, servers, WAN), or any other division that makes sense for the network. Subnetting can reduce overall network traffic and improve network performance.
Note: A subnet is equivalent to a network and these terms can be used interchangeably. Most networks are a subnet of some larger address block.