How the Year Will Play Out
I’ve run analyses and made predictions about RIR runnout, most recently in my End of 2013 blog post. So how will 2014 play? Here’s a future year in review.
Life continues as usual through January and February. IPv6 deployment increases, maybe another country has their largest ISP roll out IPv6. IPv6 is a topic at NANOG in early February, including a few positive surprises (and maybe one negative one).
Fears of exhaustion will cause accelerating IPv4 requests in LACNIC, and by March they have gone below their last /8. Crossing that threshold will drive technical headlines, and further accelerate requests, and they will hit their last /9 by end of March or maybe April.
Mainstream reporters start writing about the 8 April Windows XP End of Support. This will drive some replacement of WinXP, but most people will wait and see (they will regret that decision).
On the 8th, Microsoft issues its last update for Windows XP. On the 9th, the world experiences significant problems as 25% of the computers in the world are infected with viruses and botnets. Home electronics stores make a fortune in new computers, and the stocks of all anti-virus and security companies rise. By the 10th, there has been a slight spike in IPv6 deployment numbers, as the IPv6-incapable PCs are retired. Anti-virus vendors scramble to update their IPv6 support. Apple stock rises, Microsoft sells some devices, and headlines declare the age of the traditional home computer to be over.
LACNIC hits its last /9, triggering IANA’s Allocation Post Exhaustion policy. Under this policy, IANA allocates 1/5 of whatever fragments or returned space it has, to each RIR. Geoff Huston of APNIC estimated that reserve at about 20M addresses. So each RIR would receive about 4M addresses, the rough equivalent of a /10.
This may happen during the ARIN meeting, 13-16 April in Chicago, or the LACNIC meeting 4-9 May in Cancun. Likeliest, if the organizations can coordinate it, would be at the ICANN meeting, 23-27 March, in Singapore. Whichever meeting it is will be well-attended by trade press, and coverage of IPv4 runout will get more mainstream. Reporters may begin outlining their stories now.
IPv4 exhaustion and IPv6 will be discussed at the Brazil meeting on Internet governance 23-24 April, but nothing of substance will happen.
Following the IANA allocation, LACNIC will be under a /8. ARIN will have been right around a /8, but now has an additional /10. (These are equivalents, remember, not necessarily contiguous blocks). IANA may actually only allocate /11s, depending on the blocks they have, and distribute the rest in September (read the details of their policy closely). If that happens, LACNIC will be in the strange position of having no IPv4 addresses for a few months, then suddenly assigning a /11 in September, and being out again.
On 6 June, technologists will signal their support for IPv6 by disabling IPv4 for themselves. Following IANA’s last distribution, this event will have increased visibility, but will still not be a major event (probably even inside technical circles). Some web sites may launch their IPv6 support on this day, but it will not be a coordinated event like World IPv6 Launch. The event will be the buzz of NANOG, 2-4 June in Bellevue.
ISPs will have been deploying more IPv6 all year. The ISPs with the largest deployments will have finished rollouts, but their numbers will keep gradually rising as end users replace equipment. Global deployment per Google’s statistics will be greater than 5%, which will surprise pessimists since we started the year at 2.8%. China, in particular, will have noticeable deployments among its top three operators, and will affect the global numbers. Deployment in the U.S. will break 10% in July, as more large ISPs are rolling out IPv6, in addition to Comcast, AT&T, and TWC. Mobile carriers Verizon Wireless and T-Mobile also push the U.S. count higher. Maybe by this time Verizon FiOS and some Canadian ISPs will come out of the closet and reveal that they have been preparing for IPv6 for a long time, and do major releases in very short periods of time.
LACNIC will run out of IPv4 addresses. IPv6 deployment in the region will have surged, though CGN deployment will be even greater. Capital is scarce in Latin America, so replacing modems and CPE is very expensive compared to incremental CGN deployment.
While much of the Internet will be on holiday, a few U.S. government agencies will deploy IPv6 to their desktops, since the roadmap requires that all client applications be able to use native IPv6 to reach IPv6-enabled Internet services.
Despite requests, no government agency will yet require all contractors to use IPv6, and nobody will be using local policies to give IPv6 a stronger head start than the default behavior of Happy Eyeballs.
College campuses around the world will finish their IPv6 deployment before students return. Many of them already have huge penetration, and remaining CIOs will realize that IPv4 will have run out before the school year. They will be planning support for university servers, and pressuring their application vendors (such as Blackboard, and various MOOCs) to support IPv6.
Another noticeable surge in IPv6 may occur in September, as students move and turn up new service, and USG agencies scramble to finish their deployments. If IANA has any remaining IPv4 addresses, they will get allocated in September. LACNIC might have a single day of IPv4 allocations, and be done.
I’ve been trying to convince CDNs and hosting companies to coordinate a day in September 2014 to activate IPv6 for all of their customers (unless they opt out). A few have said they can do it; a few have said they need more pressure to update their tools and services.
The headline for all of October is the ITU Plenipotentiary meeting, 20 October – 7 November. The ITU can vote itself any powers it wishes, and based on the outcome of the WCIT, it may be that the ITU asserts certain powers and is not supported by many member states. The worst case scenario, which is not entirely implausible, is that the ITU begins driving its own version of the Internet in nations where it is least deployed and/or most feared, which could cause some level of users and content inside and outside to be unreachable to each other.
ARIN will run through its last IPv4 addresses (even the last IANA assignment) in November or December. (My projections cluster around October-November for runout, based on linear, and 2nd-3rd order polynomial projections of 2013 allocations; the IANA allocation would delay that until February 2015 (as Geoff Huston predicts), but I think there will be a surge in the last two months, bringing it back in to Q4-2014).
A panic will ensue, with people being shocked that the event predicted for 20 years actually happened. US government agencies, possibly even Congress depending on the news out of the ITU Plenipot, will have meetings or hearings and scramble to look like they are doing something useful; they will conclude that IPv6 is essential to the future of the Internet. Most consumers will still not hear anything about it, but some will include IPv6-capable home routers on their holiday shopping lists.
CIOs will see news about runout just in time to include IPv6 in enterprise IT budgets for 2015. Fortunately, when they call their ISPs, IPv6 will be available (and may already have been turned on). Some ISPs and enterprises will scramble to enable IPv6 by late November, so that they can defensibly and defensively answer questions with, “We’re already rolling out IPv6.”
A third of U.S. users will reach IPv6-capable content over IPv6 by the end of the year. There will be the November spike in deployments in expectation of and response to ARIN runout, plus cyclical rises in consumer upgrades. You can run your own projections using Eric Vyncke’s great tool. Use the last two months of 2013 (to start low and hide the anomolies), and use a logistic (S-curve) or any curve you like.
More announcements like this one from Gandi.net will come out, saying that IPv6 will be cheaper than IPv4, either immediately, or at some defined point in the future. These may be for IPv6-only servers, as at Gandi.net, or it may be that IPv6 traffic will be billed at a lower rate per gigabyte than IPv4 (for paid peering, transit, Internet service, or for hosted/CDN/cloud services). It’s possible that investment analysts will ask on earnings calls about the effects of IPv4 runout on some businesses, especially cloud services, mobile, and residential ISPs.
The price of IPv4 addresses on the transfer market will spike, then dip a little; it will take a few months for them to stabilize. IPv4 address holders, knowing that AfriNIC is the only remaining RIR with any IPv4 addresses (and no inter-RIR transfer policy) will raise their expectation of price. There will be a brief panic among potential buyers, but they will quickly settle down, and there will not be very many transactions at the new higher prices in the very short term, as buyers figure out how to respond. In addition, address holders who had been waiting for post-ARIN runout prices will finally enter the market, resulting in a temporary glut. This market instability will run through the first half of 2015, and will directly influence IPv6 deployment decisions.
IPv6 evangelists will be insufferably smug.