There was a simple plea from John Curran, CEO of American Registry of Internet Numbering (ARIN) at the latest National Telecommunications and Information Administration (NTIA) IPv6 workshop on September 28th (check the meeting agenda): “Internet is running out of IPv4 addresses.” As a matter of fact he predicted May 29th, 2011 as the end of available IPv4 allocations from Internet Assigned Numbers Authority (IANA) to Internet Registries such as ARIN. Current predictions show that individual registries will start running out of addresses by Oct. 23rd, 2011 and the last available IPv4 address will be allocated before Sep. 9th 2012. Since these dates are computed frequently (daily), the exact date is not very important. Essential thing to recognize is within the next two years, all available IPv4 addresses will be allocated.
NTIA IPv6 meeting was organized by the Aneesh Chopra (CTO) and Vivek Kundra (CIO) of the Obama administration. After the meeting, US government published a number of deadlines for the government agencies to meet. The most critical one was the completion of IPv6 capability for entire internal infrastructure by September 30th, 2014, the end of fiscal year 2013-2014. An earlier target of September 30th 2012 was set for every agency to have an IPv6 website up and running. Considering many Internet powerhouses such as Google (the earliest entrant in 2008), Facebook, Comcast have just recently managed to bring up IPv6 content, matching 2012 deadline is a formidable task for the Federal Government.
IPv6 became an Internet Engineering Task Force (IETF) Request For Comment (RFC) (RFC-1883) in December 1995. Unfortunately after 15 years, adoption of IPv6 is still quite dismal. An excellent webpage maintained by Mike Leber (Hurricane Electric) provides the summary of IPv6 adoption using different metrics. Following table is a summary of usage (as of Oct. 10th 2010) of IPv6 in Top Level Domain (TLD) nameservers for some of the most popular gTLD.
Another excellent IPv6 adoption monitoring webpage is maintained by Alain Durand (Comcast). His analysis includes top 1M domains (as reported by Alexa). Following figure shows the change in adoption during the last 12 months.
As expectedly the IPv6 adoption is significantly better among the top domains. Following figure from the same Comcast website shows the adoption for top 10, 100, 1K, etc. domains.
Unfortunately the concerning trend observed for top 1M domains seems to be valid for top 10, 100 or 1K as well: There is no increase in IPv6 adoption. Less than 2% of top 1000 domains provide IPv6 connectivity after 15 years of standardization.
Considering there are over 4 Billion customers are served by mobile network operators (MNO) and the adoption of always-connected mobile devices is skyrocketing across the world, one would have expected that MNOs would be the first adopting IPv6 and promoting the proliferation of IPv6 content. Looking back at the history of mobile standards development, we can remember that MNOs embraced IPv6 very early on. As a matter of fact, when first version of UMTS standards were developed as Release-99, IPv6 was the default choice for device addressing. Unfortunately, due to lack of IPv6 content (feels like not much has changed) as well as lack of tools, knowledge within the operator community, IPv6 was quickly dropped in favor of Network Address Translation (NAT) based solutions. This helped MNOs to operate large scale NAT solutions very early on. At a recent Google organized IPv6 implementors conference in June 2010, Verizon, T-Mobile and Tus-Mobile (from Slovenia) shared their experiences and plans for IPv6.
Comparing MNO experiences to plans of cable companies such as Comcast, Cox, Time Warner (shared at the last week’s North American Network Operator Group (NANOG) meeting), it is quite striking to see the contrast. As expected, MNOs are planning to continue to use NAT heavily:
- Those who favor dual-stack in user devices, plan to continue using NAT44 for IPv4 content while using IPv6 for native content.
- Those who plan to use IPv6 only in user devices, plan to use NAT64 (along with DNS64) for IPv4 content while using IPv6 for native content.
On the other hand, many cable providers are looking first time ever at the necessity of using NAT for IPv4. Some of these operators plan to use solutions as drastic as NAT4444 (double-NATting, at the home gateway and at their network boundary). Naturally they are concerned about scalability, cost, reliability impacts and potential difficulty to track IP address-customer mapping for external content. Amazingly, these are all the problems MNOs have been experiencing for years while trying to serve a rapidly growing subscriber base.
Depletion of public IPv4 addressing has finally made it a problem for all operators, large, small, fixed, mobile, etc. It is essential for all operators, especially MNOs to grab the opportunity, identify the best transition path and clear any roadblocks in the way of implementation. Even after 15 years of standardization, there are missing pieces in the implementation of IPv6 in many popular Operating Systems. As an example, interaction between OS and modem connection managers is an area that needs careful testing when Verizon launches its LTE network with LG USB modem in the next couple of weeks. Handset integration will be an even longer process considering very limited IPv6 capabilities in popular handset OS. A much bigger effort would be to review all the processes and systems of the MNO such that there are no roadblocks, such as provisioning or billing systems to deployment of devices with IPv6 capabilities. Considering the alternative of not having a functioning Internet anymore, I am sure that all MNOs will accelerate their efforts of deploying IPv6. In this blog, I will continue to publish IPv6 adoption status periodically as well as start adding mobile operator IPv6 deployment progress.