IPv6 deployment is essential for the continued growth of the Internet and to preserve important architectural properties, such as universal connectivity and end-to-end addressability, that have made the Internet successful in enabling a wide range of innovative applications and services.
The hope was that most computers and networks would by this time have both IPv4 and IPv6 capability, enabling a gradual transition to IPv6. Since this "dual-stack" model has failed to materialize, a number of highly undesirable scenarios are developing, or may soon. We're already seeing the rise of IPv4 address transfer markets, either sanctioned by registries or unsanctioned "black markets" where companies with IPv4 addresses to spare sell to the highest bidder. This does not bode well for the potential of the Internet as a major democratizing force in the world, as richer and more powerful entities consume scarce IPv4 addresses at the expense of others. Among the most notable recent address transfers: Microsoft's acquisition of more than 600,000 addresses from Nortel for $7.5 million in March 2011 and Borders Group selling its block of 65,536 addresses to Cerna for $786,432. And as IPv4 addresses get scarcer, prices will undoubtedly go up.
In any case, address-transfer markets are at best a stopgap.
Other adverse consequences of a lack of v6 adoption include the likelihood that service providers and enterprises will deploy more and more Network Address Translation layers, with an increasingly damaging impact on a variety of applications--and on our ability to effectively manage and monitor the network infrastructure. Severely compounding these problems will be the emergence of Carrier Grade NAT, a form of large-scale NAT that is located in the center of an Internet service provider network (rather than at the customer periphery) and that shares individual global IPv4 addresses across many customers. This combination of CGN instances, application layer gateways, and other "middleboxes" could not only impede effective network communication, it could provide a perverse mechanism enabling some service providers to lock their customers in walled-garden networks and a strictly constrained set of applications.
Finally, it's inevitable that new organizations and services will come online using IPv6 exclusively. What could then emerge is a "balkanization" of the Internet--islands of IPv4- and IPv6-only devices that will be unable to communicate. Even organizations that think they have an adequate number of IPv4 addresses will face problems, because they may no longer be able to reach this new class of emerging IPv6-only services and users. Systems are being devised to allow subsets of IPv4 and IPv6 islands to communicate (see NAT64 and DNS64), but they carry significant financial and operational costs.
World IPv6 Launch: A Success Story
World IPv6 Launch day, held on June 6, challenged organizations to permanently enable IPv6. By all accounts, it was a success, with various measurement studies showing substantial growth in IPv6 traffic. Roughly 3,000 organizations signed up to participate; the list includes top Internet sites and ISPs, including Google, Facebook, YouTube, Yahoo, Cisco, Microsoft Bing, AOL, Netflix, Comcast, Free (France), KDDI (Japan), Verizon Wireless, AT&T, and Sprint. Leading universities (Yale, Penn, NYU, Indiana, UNC), hosting providers, regional Internet registries, and standards organizations (IETF, IANA) also took part.
In terms of total volume, Free Telecom of France led, with 17% of its traffic constituting IPv6, followed by AT&T in second place. Some university networks, though, saw between 30% and 60% of traffic to their sites using IPv6--not surprising since the research and education communities have often been at the forefront of IPv6 deployment.
Google reports that IPv6 traffic has grown 150% since last year, with a 75% increase since World IPv6 Launch. Most Google services (Search, Gmail, YouTube, Google+) are available over IPv6. Facebook says that 27 million people are now using IPv6 to access its site. These companies are taking a somewhat cautious approach to maintain a high-quality experience. They make ongoing measurements of the quality of IPv6 connections from client locations, and networks that offer substantially poor IPv6 connectivity to their users are screened out from receiving DNS AAAA records; that forces those users to continue using IPv4. Thankfully, this "AAAA blacklisting" method appears to be affecting only a small number of sites, indicating that, by and large, IPv6 deployment is going well.
Large ISPs have offered IPv6 service to their enterprise customers for some time. By contrast, residential broadband has conspicuously lagged, but even this appears to be changing, especially in the United States. Cable and network service provider Comcast turned on IPv6 for 2% of its customer base and reports that IPv6 traffic has seen a 3.75% increase since last year. AT&T has enabled 6.7% of its consumer broadband customers with IPv6, which translates into more than 1 million customers. Time Warner Cable is offering trial IPv6 service in many areas of the country, too.
Wireless provider T-Mobile now has IPv6 deployed throughout its network. Verizon Wireless also provides IPv6 service on its 4G/LTE network.
Akamai, one of the world's leading content delivery networks, reports a hundredfold increase in IPv6 traffic over its trial run last year. In April, Akamai announced production IPv6 support in its network. Other CDNs, including Limelight and CloudFlare, also support IPv6.
In terms of total traffic, though, IPv6 remains miniscule compared with IPv4. Arbor Networks reports that IPv6 now accounts for just 0.15% of the total traffic across the ISP networks it monitors. This does, however, represent a twentyfold growth rate over the past year. And perhaps more importantly, the increased traffic volumes continued rather than falling back substantially as they did following last year's World IPv6 Day. Another optimistic sign is that participants reported no major operational or customer support issues.
Still, challenges remain.
For example, IPv6 has always suffered from the classic "chicken and egg" problem: Content and application service providers don't feel any urgency to deploy it until there's broad utilization in ISP networks and end sites. End sites don't feel the need until content and access networks are enabled. As IPv4 address depletion approaches a critical point, this problem may finally be overcome.
It's also true that technical problems persist. Many folks have security concerns because of lack of familiarity with the protocol; the state of maturity in IPv6 implementations in a variety of security appliances (firewalls, VPN concentrators, IDS, IPS); and the absence of high-quality DNS blocklists, used by many email abuse-prevention systems and geolocation databases. More esoteric issues include the fact that the protocol-engineering community is still working on designs for scalable multihoming, something that needs to be solved before global-scale IPv6 deployment.
None of these should be considered showstoppers. We can no longer wait for perfect solutions to every outstanding technical problem. Wider deployment of IPv6 will bring the most pressing issues to the fore quickly and solutions will inevitably emerge, even as organizations that drag their feet will ultimately pay significantly higher deployment costs with real business impacts.
The Future Is IPv6
IPv6 is on by default in all modern computer operating systems. CIOs, if they haven't already, should work on a strategy for IPv6 adoption, deployment, and migration. Obtain IPv6 address blocks from your ISP or regional Internet registry; train your IT staff; work with ISPs on enabling IPv6 network connectivity; and begin to enable IPv6 in websites, DNS, email systems, and other application services. This requires careful planning and execution, so it's wise to start as soon as possible.
Perhaps Vint Cerf, widely considered one of the fathers of the Internet, sums it up best: "There are no more excuses. You have to be able to run both IPv4 and IPv6 all the time, anytime, because this is not going to be turned off. For any ISP and any edge provider, if you are not capable of running IPv6, you are not being noticed. Get going. Get IPv6 running."
Shumon Huque is the director of engineering, research and development for the University of Pennsylvania's Networking & Telecommunications division and also serves as the lead engineer for the MAGPI GigaPoP.