There are three classes of address space used with IPv4, the current version of IP:
Class A networks are commonly referred to as "slash-eight" networks because they use an eight-bit network prefix. Each Class A network can support 16.7 million computers. There are 128 Class A networks. About half of all possible IPv4 addresses - more than 2.1 billion - reside within Class A. Class A holders include U.S. military agencies, research institutions involved in creating the Internet and large corporations that were smart enough to get in on the Internet boom early.
Class B networks are called "slash-sixteens" because they have a 16-bit network prefix. Each Class B network can support up to 65,534 hosts. There are 16,000 Class B networks. A quarter of the IPv4 address space - representing more than one billion addresses - reside in Class B. Most Class B holders are large network customers and ISPs.
Class C networks are called "slash-twenty-fours" because they have a 24-bit network prefix. Each Class C network can support 254 hosts. There are two million Class C networks. The entire Class C space contains 536 million addresses, representing about an eighth of the total IPv4 address space. Most Class C holders are ISPs and midsize businesses.
Two additional classes exist: Class D for IP Multicast traffic and Class E for experimentation.
In the early 1990s, the Internet engineering community realized that these classes were not an efficient means of doling out IPv4 addresses. Class A blocks were larger than any organization could possibly use. Most enterprise customers required an address space that fell between Classes B and C.
As a result, the community developed a routing technique called Classless Inter-Domain Routing (CIDR), which supports arbitrarily sized blocks of network addresses rather than the standardized blocks used by Classes A, B and C.
Unused portions of all three original classes can be reassigned as CIDR blocks of varying sizes.
At the same time, the Internet engineering community began recommending that organizations without enough IPv4 addresses should create private Internet addresses. Using a technique called network address translation (NAT), an organization can assign a single public IPv4 address to multiple private Internet addresses. Under this scenario, IPv4 addresses are retained for host computers that require external communications with the Internet. NAT technology, however, makes it harder to deploy certain network applications and security mechanisms that require end-to-end addressing.