IP addresses and subnetting

Every device on an IP network needs an address, and every router makes its decisions from addresses. To design networks, configure cloud environments or read a firewall rule, you need to see an address the way a router does: as a string of bits with a boundary in the middle.

An address is 32 bits

An IPv4 address such as 10.20.30.77 is really a 32-bit number. The dotted notation just writes it as four 8-bit chunks, called octets, each shown in decimal from 0 to 255.

10       .20       .30       .77
00001010 .00010100 .00011110 .01001101

32 bits gives about 4.3 billion possible addresses — which the world ran out of years ago. That shortage is why private addresses, NAT and IPv6 exist, all of which you will meet in this course.

Network part and host part

Every address is split into two parts:

• the network part — which network the device is on (shared by everyone on the same subnet);
• the host part — which device it is within that network.

Where the split falls is given by the prefix length, written after a slash in CIDR notation. In 10.20.30.77/26, the first 26 bits are the network part and the remaining 6 bits identify the host.

The same boundary can be written as a subnet mask: 26 ones followed by 6 zeros, which in dotted form is 255.255.255.192. Older equipment and many settings screens still ask for the mask; modern tools prefer the prefix length. They mean exactly the same thing.

Try the calculator

Change the address and the prefix length and watch which bits belong to the network (teal) and which to the host (amber). Then test a second address to see whether a device would deliver to it directly or hand it to the router.

Working out a subnet by hand

The calculator does four simple things you can do yourself:

1. Network address — keep the network bits, set every host bit to 0. For 10.20.30.77/26 that is 10.20.30.64.
2. Broadcast address — keep the network bits, set every host bit to 1: 10.20.30.127. A message sent here reaches every host on the subnet.
3. Size — a prefix of n leaves 32 − n host bits, so the subnet holds $2^{32-n}$ addresses. Here, $2^6 = 64$.
4. Usable hosts — subtract the network and broadcast addresses: 62 hosts, from .65 to .126.

A useful shortcut: in the last octet, subnets of a given size always start at multiples of that size. /26 subnets (size 64) start at .0, .64, .128 and .192 — so .77 must fall in the one starting at .64.

Local or remote? How a device decides

Before sending anything, a device compares its own network part with the destination’s, using a bitwise AND with the mask:

• Same network → the destination is on the local link. The device finds its hardware address (using ARP on IPv4) and delivers the frame directly.
• Different network → the device sends the packet to its default gateway, the router, which forwards it onward.

This is why a wrong mask or gateway causes such confusing faults: a machine may try to deliver “locally” to an address that is actually remote, or send local traffic to a router that drops it.

Private and special ranges

Some ranges are reserved and never routed on the public internet. Organizations reuse them freely inside their own networks:

10.0.0.0/8

Private — about 16.7 million addresses. Common in large enterprises and cloud networks.

172.16.0.0/12

Private — 172.16.0.0 to 172.31.255.255.

192.168.0.0/16

Private — typical of home and small-office routers.

127.0.0.0/8

Loopback — 127.0.0.1 always means “this machine”.

169.254.0.0/16

Link-local — self-assigned when no DHCP server answered. Seeing one usually signals a problem.

Subnets tell a device when to hand traffic to a router. Next, we follow what the router does with it.