Subnetting a network into equal subnets

Why subnet at all?

Imagine an office with 600 PCs, printers, cameras and phones, all in one big network. Every broadcast (a message to everyone, such as an ARP request asking "who has this IP?" or a DHCP Discover) reaches all 600 devices. One misbehaving device can flood everyone. Accounts and the guest Wi-Fi sit side by side with nothing between them. And when something breaks, nobody knows where to look.

Subnetting splits that big network into smaller networks called subnets. Each subnet is its own broadcast domain: a broadcast stays inside it. A router (or a layer 3 switch) connects the subnets, and that is where you can put security rules. Subnets also give structure: one per floor, per department or per device type.

Borrowing bits

A network has a fixed number of host bits. To make subnets, you borrow some of those host bits and use them to number the subnets. The prefix gets longer by the number of borrowed bits. Start with 192.168.1.0/24: 8 host bits. Borrow 2 bits and the prefix becomes /26, with 6 host bits left.

• Number of subnets = 2^s, where s is the number of borrowed bits.
• Usable hosts per subnet = 2^h − 2, where h is the number of host bits left.
• s + h always equals the host bits you started with.

Worked example 1: 192.168.1.0/24 into 4 subnets

We need 4 subnets. 2^2 = 4, so borrow 2 bits. New prefix /24 + 2 = /26, mask 255.255.255.192. Host bits left 6, so 62 hosts each. Block size 256 − 192 = 64, so the subnets start at 0, 64, 128 and 192.

Worked example 2: the same /24 into 8 subnets

2^3 = 8, so borrow 3 bits. New prefix /27, mask 255.255.255.224, block size 32, hosts 2^5 − 2 = 30 each. The networks are 192.168.1.0, .32, .64, .96, .128, .160, .192 and .224, all /27. The last one runs from .224 to .255, with hosts .225 to .254.

Worked example 3: 172.16.0.0/16 into /20 subnets

Going from /16 to /20 borrows 4 bits, so there are 2^4 = 16 subnets. Host bits left 12, so 4,094 hosts each. The mask is 255.255.240.0 and the block of 16 is in the third octet. The subnets are 172.16.0.0/20, 172.16.16.0/20, 172.16.32.0/20 and so on up to 172.16.240.0/20. The first one runs from 172.16.0.1 to 172.16.15.254 (broadcast 172.16.15.255). The last one runs from 172.16.240.1 to 172.16.255.254 (broadcast 172.16.255.255).

Working from host counts

Often the question is not "how many subnets" but "each subnet needs H hosts". Find the smallest h where 2^h − 2 is at least H. For 50 hosts: 2^5 − 2 = 30 is too small, 2^6 − 2 = 62 is enough, so h = 6 and the prefix is 32 − 6 = /26.

import ipaddress

net = ipaddress.ip_network("192.168.1.0/24")
for subnet in net.subnets(new_prefix=26):
    print(subnet, subnet.broadcast_address)

Common mistakes

• Subtracting 2 from the number of subnets. Older textbooks did this because the first and last subnet ("subnet zero" and the all-ones subnet) were once avoided. Cisco IOS has allowed subnet zero by default since IOS 12.0, and the CCNA uses 2^s.
• Borrowing too many bits: always check that the subnets you get still hold enough hosts.
• Starting a subnet at a number that is not a multiple of the block size, such as 192.168.1.50/26. It is not a valid network address.