88 IP Addressing in the Global Internet
By Andrew McConachie, December 2013.
Overview. Most people take for granted that the Internet just works. They connect their computer to the Internet, it gets an IP address, and they are able to communicate with a computer with a different IP address on the other side of the planet. How did their computer get the correct IP address? How does any computer or router get the correct IP address? How did the routers and other computers on the Internet get their IP addresses? Who decides which computers and which routers get which IP addresses?
What is being organized? At their simplest, an IPv4 address is a 32-bit series of 0′s and 1′s. They are resources that are born-digital, as they have no canonical physical representation. Their digital canonical representation, with which we have all become familiar, is called the “dotted quad” format and is 4 numbers between 0-255 separated with dots. For example, 169.229.216.200 is the IPv4 address for www.berkeley.edu.
Not all IP addresses are of equivalent classes. There are unicast, multicast, broadcast, and experimental IPv4 addresses, and unicast addresses can be either public or private. There are also two different versions of IP addresses currently in use on the Internet, IPv4 and IPv6. We will focus on IPv4 unicast public IP addresses, since these are not only the most common, but also the most important. This is roughly the range of IP addresses from 1.0.0.0 to 223.255.255.255, with some breaks in the middle for private IP address space.
Why is it being organized? IP addresses are the foundation of network connectivity and the Internet; they identify each device on a computer network and also serve as its address, so that routers and other devices can locate and communicate with it. You cannot get online without one. IP addresses can be represented into blocks, or subnetworks, using a prefix and a mask. For example, 169.229.216.0/24 represents all IP addresses in the range of 169.229.216.1 – 169.229.216.255. Internet routers do not have enough memory to hold routes for every individual IP address on the Internet. So by organizing the Internet into subnetworks based loosely on a hierarchical model, routers are able to determine the correct path for every destination in the network without actually storing every address in their memory. If the organization of IP addresses is not handled properly, Internet routers would exhaust their memory space and parts of the Internet would become unreachable.
How much is it being organized? Currently there is too much granularity in the global Internet routing table. For a router it takes the same amount of memory to store a subnetwork with 255 IP addresses as it does to store a subnetwork with 65536 addresses. So if our main concern is to minimize memory usage in Internet routers, thereby lowering operator costs and increasing stability, we want as little granularity as possible in the Internet routing table. The problem is that many organizations use non-contiguous IP subnetworks that cannot be aggregated into larger subnetworks. This results in routers having to store many small subnetworks instead of fewer larger subnetworks, which will eventually lead to memory exhaustion in older routers and possible reachability issues. Currently the full Internet routing table is approaching 500,000 routes. Most network engineers expect problems once the routing table grows past 512,000 entries, since router memory limitations are always at bit boundaries.
When is it being organized? IP addresses are organized once someone configures one on a device or sets up a Dynamic Host Configuration Protocol(DHCP) server. If an organization exhausts their supply of free IP addresses. it will have to make a request to the upstream provider or Regional Internet Registry(RIR) for more address space. In the early days of the Internet, large blocks of IP addresses were given to organizations, but this led to many of the addresses in these blocks not being used. We are now reaching a point where we no longer have new addresses to assign to organizations.
Markets are now emerging for organizations to buy and sell IP addresses, and the organizations who have held on to large amounts of unused addressing space stand to make significant revenue from selling their unused space. When these organizations sell their unused IP address space, they will break up large allocations into smaller subnetworks, thereby increasing granularity and further accelerating the growth of the Internet routing table.
How or by whom is it being organized? The Internet Corporation for Assigned Names and Numbers(ICANN) is currently responsible for initial allocation of IP addresses. They allocate 8 blocks of IP addresses to RIRs, who are then responsible for distributing allocations to organizations that request them. These organizations can then allocate IP addresses to smaller organizations, thus forming a loose hierarchy of organizations, where each level lower in the hierarchy receives a subset of the IP address space from the organization above it. ICANN no longer has any /8 blocks of IP addresses left to allocate to RIRs. Once all of the RIRs have exhausted their last allocations from ICANN, organizations will have to rely on secondary markets to increase their IP address space.
Other Considerations. The world of IP addressing will soon get a lot more interesting. The introduction of IPv6 as a replacement for IPv4 has been slow in coming and, while gathering momentum, continues at a snail’s pace. As organizations start purchasing IP addresses from one another, we should expect increased granularity and decreased stability in the Internet routing infrastructure. Whether or not normal Internet users notice will ultimately be determined by how well equipment vendors and engineers expediently address the coming problems.
