Last month, the Obama administration sponsored one of the first high-level government workshops on IPv6. At the meeting, the administration’s Chief Information Officer, Vivek Kundra, announced a remarkable 2012 deadline for federal agencies to support IPv6.
So with a high-level US government mandate and a recent spade of vendor and carrier IPv6 announcements (e.g. Verisign, Comcast and Hurricane Electric), is the 15 year old IPv6 migration effort finally gaining momentum?
As discussed in earlier blog posts (see “The End is Near, but is IPv6?”), v6 Internet deployment has lagged even some of the most pessimistic predictions of a decade earlier. Our 2008 Technical Report found minuscule levels of IPv6 traffic — less than 1/100 of one percent of Internet traffic. The tech report also noted a near complete lack native IPv6 traffic measurement infrastructure. None of the 110 ISPs participating in ATLAS in 2008 had native v6 measurement capability (i.e. either no native v6 or routers lacking v6 flow export capabilities).
On the positive side, five anonymous ATLAS deployments began sharing native IPv6 traffic statistics for the first time this month. This represents a significant milestone in ATLAS IPv6 visibility and our third annual analysis of Internet v6 traffic. The below table shows the average percentage of native, 6to4 and total v6 traffic in five carriers (two in North American and three in Asia / Europe).
Interestingly, 6to4 traffics runs 5x or higher than native IPv6. This may be because the five ATLAS networks lack complete v6 backbone router instrumentation or, equally likely, thousands of end-users still rely on tunnels to connect their home or enterprise v6 islands to the Internet (i.e. CPE devices or enterprise networks lack native v6 support). Indeed, even looking at small regional ISPs, the data shows hundreds of long-lived 6to4 tunnels across 70 percent or more of the customer base. Clearly, there is some demand for end-user v6 connectivity.
But Internet wide IPv6 measurements still show relatively little v6 traffic. The graph below shows 6to4 as an average percentage of all Internet traffic across ATLAS ISPs. While v6 traffic has climbed dramatically in the last three years (more than a hundredfold), v6 remains less than one twentieth of one percent all Internet traffic as of October 2010. If you’re curious, I discuss some of major sources of v6 growth in earlier blogs and some of the smaller spikes correspond to v6 only Internet engineering meeting sessions.
At this rate, we have years to go before even modest levels of v6 adoption.
Still, I remain a cautiously optimistic (or maybe that’s slightly pessimistic) proponent of IPv6. Although IPv6 adoption remains painfully slow, as this in-depth Ars Technica article argues, we lack a plan B — the IPv4 address registry exhaustion deadline looms. We need to make IPv6 work.
Unfortunately, technical hurdles like poor DNS implementations, lack of IPv4 parity features in protocol standards (e.g. DHCP), missing vendor support, and a rats nest of tens of thousands of poorly configured tunnels all raise the cost, degrade the quality and otherwise delay broader IPv6 adoption. Personally, I tend to think we will not see significant v6 adoption until a vibrant IPv4 market exists with corresponding IPv6 economic incentives. Although controversial, the liberalized ARIN transfer policy is a step in the right direction.
As Ram Mohan notes in his CircleID blog post, a lack of enterprise demand and education are probably some of the biggest obstacles to IPv6 adoption. But also on the positive side, the last several years have seen a growth in IPv6 outreach (e.g. ARIN and Hurricane Electric have done a great job) as well as an expanding selection of commercial enterprise IPv6 training courses.
The full page magazine advertisement below shows one good example of v6 training. The ad offers a day long seminar of “topics you don’t want to miss”, including hands-on v6 instruction and an in-depth look at everything you need to know for the imminent IPv6 transition.
So maybe we’re building IPv6 momentum. Maybe the more than 15 years of development effort, carrier testing and thousands of hours of IETF discussion and IPv6 outreach will pay off. Maybe consumer and enterprise IPv6 demand will grow before we exhaust IPv4 allocations.
Unfortunately, you are too late for the above Network World IPv6 seminar.
This training course took place in 1996.