Subnetting and CIDR Basics — Understand IP Address Division from Scratch

"So how many addresses does a /24 actually give me?" "What is the difference between a subnet mask and CIDR?" — these are common stumbling blocks in network design. In this article we start from the 32-bit structure of an IPv4 address and build up to how the network and host portions are split, subnet masks and prefix length, the math of usable host counts, and a step-by-step example of actually splitting a subnet — all with verified, concrete examples.

1. What an IPv4 address is — 32 bits and four octets

An IPv4 address is a 32-bit number. We divide it into four groups of 8 bits, write each as a decimal number (0–255), and join them with . to get the familiar notation. Each 8-bit group is called an octet.

For example, 192.168.10.1 looks like this in binary.

One octet is 8 bits, so it can represent 2^8 = 256 values, that is 0–255. Across all 32 bits you can express 2^32 ≈ 4.3 billion addresses.

2. The network portion and the host portion

An IP address is broadly split into two parts. The leading network portion tells you "which network (subnet) this is," and the trailing host portion tells you "which device within that network." As a postal analogy, the network portion is the "town and street," and the host portion is the "room number" within that address.

Devices that share the same network portion can communicate directly without going through a router. Conversely, to reach a host with a different network portion, traffic goes through a router. In other words, deciding "how far the network portion extends" is exactly what it means to split (subnet) a network.

3. Subnet masks and prefix length (CIDR)

What tells you "how far the network portion extends" is the subnet mask. The mask is 32 bits, and it marks the network portion with 1s and the host portion with 0s. The 1s are always packed contiguously from the left.

For instance, if the network portion is the leading 24 bits, the mask looks like this.

CIDR notation (Classless Inter-Domain Routing) is a shorthand that writes the number of consecutive leading 1s in the mask as a slash and a number, such as /24. 255.255.255.0 and /24 mean exactly the same thing. For example, 192.168.10.1/24 means "the address 192.168.10.1, with the leading 24 bits as the network portion."

4. Network address, broadcast address, and usable host count

Within a single subnet, two addresses carry special meaning.

• Network address: the address where the host portion is all 0s. It is the "name" of the subnet itself and is not assigned to any device.
• Broadcast address: the address where the host portion is all 1s. It is used to send to every device in the subnet at once, and it too is not assigned to any device.

Therefore, in a subnet with prefix length n, the usable host count that can actually be assigned to devices is given by the following formula.

Take 192.168.10.0/24 as an example. The host portion is 8 bits, so the total number of addresses is 2^8 = 256, and the usable hosts number 256 − 2 = 254. Concretely, it works out as follows.

The /31 and /32 exceptions

The − 2 rule above has exceptions.

• /31: there are only 2 addresses (2^1), so by the formula the usable host count would be 0. To handle this, RFC3021 specifies that for point-to-point links between two routers, both addresses may be used as hosts. No network or broadcast address is set aside.
• /32: there is just 1 address, used to point at a specific single host. It is used for loopback addresses, or in routing and access control to specify "this one device only."

5. Private addresses (RFC1918) and IP classes

Separate from the global addresses used on the internet, private addresses are reserved for use inside an organization's network. RFC1918 defines the following three ranges. On home or office LANs, you can use these ranges freely.

Because these addresses are not used on the internet, they are translated to a global address by NAT when communicating externally.

The former IP classes (A / B / C)

Before CIDR became widespread, the class was fixed by the leading bits of the address. Knowing this as historical background also makes the "old class" column in the table above easier to understand.

6. A subnetting example — splitting a /24 into four /26 subnets

Suppose you want to split one /24 (254 hosts) into smaller pieces, one per department. By borrowing 2 bits from the host portion and giving them to the network portion, the prefix length becomes 24 + 2 = /26, and the subnet is divided into 2^2 = 4 pieces.

A /26 has a host portion of 32 − 26 = 6 bits, so the total number of addresses per subnet is 2^6 = 64, with 64 − 2 = 62 usable hosts. Splitting 192.168.10.0/24 into four yields the following.

Notice that the network addresses go up by 64 each time: 0, 64, 128, 192. This matches the total number of addresses in one subnet (64). This "increment (block size)" is given by 2^(32−n) and is the single most useful landmark when doing subnet calculations.

Common prefixes and host counts