Guest post written by David Conrad
David Conrad is a technical engineer at CloudFlare, a provide of Web content optimization software.
By most any measure, the Internet has been wildly successful. From its beginnings over 40 years ago, the Internet has become a globe-spanning infrastructure upon which national economies now depend. However, that very success has led the Internet to a crossroads and the path the Internet takes moving forward will have an impact on how the network continues to grow and its openness to innovation.
To help explain the situation, imagine a city that is growing so rapidly that all the telephone numbers in the city’s single area code are nearly used up. There are four ways telephone companies can handle this problem:
- Stop providing new telephone service;
- Get people to use private extensions on their telephone lines;
- Go back and try to find telephone numbers that aren’t being used and reallocate them; and/or
- Add another area code.
Historically the most frequently chosen option has been the last. Despite the disruption new area codes cause, they allow the telephone system to meet future growth requirements.
The Internet is now facing a similar situation, with roughly similar options. The Internet’s current numbering system is limited to approximately 4.3 billion numbers. Today, the organizations that hand out those numbers are allocating around 300 million numbers, technically known as Internet Protocol version 4 or IPv4 addresses, per year. Current projections suggest the last of those addresses will be handed out within four years. At that time, the same four options, with modifications, will apply:
- The Internet can stop growing;
- “Private extensions” can be added to addresses;
- Allocated but unused addresses can be recovered; and/or
- The total number of addresses can be expanded.
The first option isn’t being considered: the Internet has become too important for growth to stop, particularly given only 2 billion of the world’s 7 billion people are connected to the Internet. The debate is whether the Internet should be upgraded to have more addresses or will a combination of private extensions and recovering unused addresses be sufficient for future connectivity needs.
The option that appears to be the most appealing in the short term would be the use of private extensions. In the Internet, something like private extensions are implemented by a set of technologies known as Network Address Translation or NAT. NAT translates a single “public” address into thousands of “private” addresses on the fly. Like connecting to a telephone switchboard, NAT acts as the operator, forwarding a call on to an internal telephone extension. Many people on the Internet today already use NAT, but these private extensions make the network much more complicated, prone to breaking, and provide convenient places where hackers can intercept network traffic. Worse, massive use of NAT, as would be necessary for the Internet to continue growing at its current rate, would encourage a move towards a world in which network providers have control over which websites are accessible to their users and which users can publish information on their websites.
Unfortunately, a NAT-facilitated world is one in which innovation would be constrained to what the network providers would allow.
Looking at recovering addresses, in his Forbes.com post “Pssst! Rare IPv4 Addresses For Sale! Get Them While You Can!”, Phil Lodico suggests that the market can be used to help extend the life of IPv4. When the Internet was first being put together, few if any believed the network would grow to its current size and little thought was given to conserving addresses. It seems logical, albeit against Internet tradition, that organizations with more addresses than they need, should be able to sell those addresses to those who need them. This buying and selling would increase address use efficiency and reduce the pressure on the remaining pools of unallocated addresses.
However, the financial considerations of a market solution are daunting, at least for address purchasers. Mr. Lodico’s article provides an estimate of 300 to 400 million “addresses available for transfer.” But at current consumption rates, this would translate into just over a year’s supply, which at the “conservative” estimate of $20/address would cost between $6 billion and $8 billion. Of course, as the supply of addresses diminishes and demand driven either by continued Internet growth or by speculation increases, the costs to obtain addresses could escalate dramatically. Today, the cost for addresses is negligible, less than a penny per address for a large numbers of addresses. Increasing costs by several orders of magnitude may reduce interest in this option. Unfortunately, to date, no other mechanism has been successful in recovering significant amounts of allocated-but-unused addresses.
The final option, expanding the total number of addresses, is most appealing. Unfortunately, this option is vastly more difficult than adding an area code in the telephone system. A closer analogy, at least in the U.S., would be adding a bunch of digits to the front of telephone numbers. All the equipment that assumes telephone numbers are 10 digits, from handsets to telephone company switches and everything in between, would have to be upgraded or replaced, telephone books reprinted, staff retrained, etc. Similarly, upgrading the Internet to support larger addresses, that is, moving from IPv4 to the new version, known as Internet Protocol version 6 or IPv6, is a massive undertaking, estimated by NIST – National Institute of Standards and Technology - in 2003 to cost $25 billion over 25 years.
However, in addition to some evolutionary improvements to the Internet’s underlying technology, IPv6 provides for 340 undecillion addresses, or 340 followed by 36 zeros, a number, it is said, sufficient to count every atom on the surface of the Earth and still have enough numbers left to count the atoms on more than 100 additional planets. Moving to IPv6 allows for essentially unlimited growth, at a reasonable cost, compared to the estimate of $6-8 billion for just one additional year of living with IPv4.
Just as the Internet disrupted the then existing data networks, providing a fertile ground for innovation, new products and services that will facilitate the transition to the vastly larger IPv6 Internet are now emerging. For example, on September 27, CloudFlare announced its Automatic IPv6 Gateway, a new free service that allows IPv4-only customers to support the IPv6-using Internet by simply turning on a switch. The same free service allows IPv4 Internet users to access websites that have already migrated to IPv6.
As the remaining pools of unused IPv4 addresses are exhausted and obtaining additional IPv4 addresses becomes difficult and costly, Internet users will see additional products and services that facilitate the transition to IPv6, allowing the Internet to continue to grow. As such, the challenge facing the Internet isn’t so much managing the transition to IPv6, rather it is avoiding the allure of short-term market-based solutions aimed at extending the life of IPv4 past its “sell-by” date.
