An IP address is defined in decimal by four groups of three digits, with each group or octet, separated by a dot (.).
Each octet in binary is a set of eight digits.

There are several classes of Internet Protocol (IP) addresses.

Each class is defined by how many network versus host addresses it supports.

Three of these IP address classes are outlined below:

Class A
Range 10.0.0.0 to 126.255.255.255
Default subnet 255.0.0.0
(127.0.0.0 is use for loopback so is not included in a class)

Class B
Range 128.0.0.0 to 191.255.255.255
Default subnet 255.255.0.0

Class C
Range 192.0.0.0 to 223.255.255.255
Default subnet 255.255.255.0

A class C address generally supports one network with 254 hosts.

The Class C default subnet mask in binary is  11111111.11111111.11111111.00000000.

The first three octets are the network address and the last octet is the host address.

The “24” CIDR notation is just a shorthand representation of the subnet mask (the first three octets have 24 bits).

Why all this is important is that it allows the network designer to create custom subnets from a single network address.

If you are using a network address of 192.168.0.0/24. This represents one network with 254 hosts (actually there are 256 different addresses with this combination, but the first is reserved for your network address and the last combination represents the broadcast address).

In the case where you wanted to create a custom subnet mask, dividing your network into two.

To do this, you have to “borrow” two bits (you can’t borrow just one (there is an exception, but that’s more advanced) from the host octet and add them to your network address for a total of 26 bits.

This would give you two networks containing 62 host addresses each (actually 64, but again, you have to reserve the first address for the network and the last for the broadcast).