Classless IP

When TCP/IP was created the original designers did not foresee the popularity of the Internet and the current move toward a networked computer on every desk and in every home. The original address design is under critical strain from the explosive growth of the Internet. The design has caused two major problems: Address Depletion and Router Table Overload. Address depletion is a very straightforward problem; the number of available addresses is being used up. Simple arithmetic shows that use of a 32-bit address translates to over four billion possible addresses. There are not yet 4 billion computers on the Internet, however; the addresses are being used up. This is because large numbers of addresses either cannot be used, or are being wasted. Using Classical IP assignment, if a network has less than 255 hosts, it is assigned a Class C. This means if the network has 2 computers, then over 250 addresses are wasted. If a network has 255 hosts, it is assigned a Class B and over 65 thousand addresses are wasted. This is a very inefficient use of addresses. Router table overload is a more immediate problem than address depletion. At the top level of the network hierarchy there are National Access Points (NAPs) where the networks of all of major Internet Service Providers (ISPs) come together. Each major ISP has one or more routers at the NAPs. The NAP routers have tables with route entries for all the networks on the Internet. In the first few years of Internet existence, the number of routing table entries was only a few hundred. With the popularity of the Internet, those tables began to grow at a geometric rate. Every time a new network was assigned, its entry was added to every top level router. This meant that the memory of every one of these routers had to be expanded at a steady rate, which was an expensive proposition. Also, there is a physical upper limit to the amount of memory that will fit in these routers. Either they would have to be replaced with newly designed (and no doubt extremely expensive) equipment, or some other solution would need to be developed. Two solutions were developed for the problems of address depletion and router table overload. The first is called IP version 6 (IPv6). IPv6 is a version of the IP protocol that uses 128 bit addresses instead of the current (IP Version 4) 32-bit addresses. A conservative estimate indicates that this scheme would provide more than 1500 addresses for every square meter of the Earth's surface. The IPv6 standard is still in development at the time of this writing. Deployment of IPv6 will probably mean the replacement of all NAP routers with newer, more expensive equipment, but this part of the problem is also under consideration. Until IPv6 can be completed, and all of the intricacies of deployment worked out, an interim solution is needed. That interim solution is called Classless IP. Classical IP divides the address space into Class A, B, and C networks by dividing the 32-bit address on 8-bit (byte) boundaries. Classless IP divides, or subnets, these classic networks into smaller networks by breaking them on bit boundaries. Classless IP also allows multiple classical networks to be combined together, or aggregated, to form a supernet. Classless IP uses a 32-bit number called a mask to define the network. The value is sometimes called a subnet mask or a netmask. Netmasks are usually written using the same dotted-quad format used in writing an IP address. The netmask is actually a 32-bit value that starts with a contiguous group of ones followed by a contiguous group of zeroes. For example, the following values are valid netmasks: Binary Dotted-Quad 11111111 11111111 11111111 00000000 255.255.255.0 11111111 11111111 00000000 00000000 255.255.0.0 11111111 00000000 00000000 00000000 255.0.0.0 11111111 11111111 11111111 11100000 255.255.255.224 11111111 11111111 11111111 11111000 255.255.255.252 11111111 11100000 00000000 00000000 255.32.0.0 11111111 11111111 11111000 00000000 255.255.252.0 The following are NOT valid netmasks: 11111111 00000000 11111111 00000000 255.0.255.0 11111111 11111111 11111111 00001111 255.255.255.15 The one bits in a netmask must be contiguous, so some software and equipment developers designate a netmask by the number of ones in the mask. For example, a netmask of 255.255.255.0 has 24 one bits followed by 8 zeroes. This is referred to as a 24 bit mask. Using this notation an address and mask pair is written with the address followed by a slash and the number of bits, such as 192.168.10.30/24. Classic IP rules assign each new organization a block of 255, 65 thousand, or 16 million hosts. Classless IP subnetting allows the assignment of a block of addresses that is much closer in size to the actual number of addresses needed on a network. For example, if an organization will only have about two dozen hosts, they can be assigned a subnet using a mask of 255.255.255.224 that will give them a block of 30 host addresses. If an organization will have about 950 hosts, rather than assign a class B network and waste over 64 thousand addresses, a network with a mask of 255.255.252.0 can be used. This will aggregate four consecutive class C networks, or subnet parts of a class A or class B network, to yield a block of about 1000 host addresses. Classless IP offers a partial solution for the problems of address depletion and router overload. By breaking up the address space into more appropriate sized pieces, classless IP cuts down on wasted addresses, allowing the available addresses to be used much more efficiently. It allows the problem of router overload to be addressed in another fashion. Under classless IP, rather than having networks assigned to individual organizations by a centralized authority, large blocks of IP addresses are assigned to the major ISPs that have routers at the NAPs. Using netmasks, a block of 1024 class C addresses uses a single routing table entry rather than 1024 individual entries. The ISP then breaks up the block inside their own network, assigning aggregated or subnetted blocks of addresses as needed. This has dramatically slowed the growth of top level routing tables, and has in some cases allowed sections of the existing routing tables to be compressed. Classless IP is a great improvement over Classical IP, but is not a permanent solution. The growth of the Internet does not appear to be slowing, and even the most aggressive efforts to conserve available addressing will probably not meet the needs of the Internet much longer. Full deployment of IPv6 or some other solution is needed soon.