All networks in practical use have different sizes. For example, a company that will have 50 computers, will not need a network of 5000 computers, And on the contrary, a company that needs 5000 computers does not need a network that can only hold 50 computers.
This is the main reason that engineers decided that IP address space should be divided in different classes in order to meet different requirements.
The following classes were created; each class fixes a boundary between the network prefix and the host within the 32-bit address:
Class A Network (/ 8 Prefixes)
This network is 8-bit network prefix. Its highest bit is set to 0, and contains a 7-bit network number and a 24-bit host number.
A maximum of 126, which is (2 7 -2,) networks can be defined; two is subtracted because all an (0 and 1) subnet cannot be used in certain routers using RIP-1 Protocol. Each network supports a maximum of 16,777,214 (2 24 -2) hosts per network. You must subtract two because the base network represents host “0”, and the last host on the network is actually used for 1s ("broadcast") and may not be assigned to any host.
The class A network address block contains 2 31 power (2,147,483,648) individual addresses. The IPv4 address space contains a maximum of 2 32 power (4,294,967,296) addresses, which mean that a class A network address space is 50% of the total IPv4 unicast, address space.
Class B Networks (/16 Prefixes)
This network is a 16-bit network prefix; its highest bit order is set to 1-0. It is a 14-bit network number with a 16-bit host number.
This class defines 16,384 (2 14 ) /16 networks, and supports a maximum of 65,534 (2 16 -2) hosts per network. Class B /16 block address is (1,073,741,824) = 2 30; therefore it represent 25% of the total IPV4.
Class C Networks (/24 Prefixes)
This is a 24-bit network prefix; it has a 3 bit set to the highest order 1-1-0. It is a 21-bit network number with 8-bit host number.
This class defines a maximum of 2,097,152 (2 21 ) /24 networks. And each network supports up to 254 (2 8 -2) hosts. The entire class C network represents 2 29 (536,870,912) addresses; therefore it is only 12.5 % of the total IPv4.
There are two other networks that are not commonly used, class D and Class E.
Class D has its highest bit order set to 1-1-1-0 it is used to support multicasting.
Class E has its highest bit order set to 1-1-1-1 which is reserved for experimental use.