When it became clear that 32-bit IP addresses just wouldn't cut it for a growing Internet, the Internet Engineering Task Force did what its name suggests and created a new version of IP. IPv6 has so many addresses that it will resist our best efforts to waste all of them for many decades, if not centuries. Of course to benefit from the larger IPv6 address space, it's necessary to actually migrate from the existing IPv4 to the new IPv6. That part has been sorely lacking in the 15 or so years since the first IPv6 specifications were published. The good people over at Arbor Networks did a study in 2007 that showed surprisingly little uptake of the new protocol. And remember, that surprise was on top of the already set-in realization that IPv6 wasn't taking the world by storm in the first place. And now, as IPv4 addresses have run out in part of the world already, Arbor has repeated its study.
The result? A reduction in tunneled IPv6 traffic during the study period.
First the good news. The 2007 study was criticized because the measurement methodology was only capable of detecting tunneled IPv6 traffic—IPv6 packets embedded in an IPv4 packet. "Native" IPv6, where the IPv6 packet is transmitted directly just like IPv4 packets, couldn't be measured at that time. Now, six networks in North America and Europe are also able to measure native IPv6.
The new study ran for six months starting late last summer, looking at a total traffic volume of about 8 terabits per second. The IPv4 traffic grew by 40 to 60 percent during the six-month period, but tunneled IPv6 traffic peaked at 0.04 percent in August and then declined by 12 percent over the course of the study. (That still adds up to some 3 Gb of tunneled IPv6 traffic.)
The study speculates about the reason for the decline, but doesn't offer a definitive conclusion. Let's fix that: Macs would use the 6to4 automatic tunneling mechanism where available in the past, but since the Mac OS X 10.6.5 update that came out last year, Macs have joined other major systems in preferring IPv4 over 6to4 IPv6. Macs have made up a large percentage of all systems using 6to4 in recent years because Apple's Airport Extreme base stations at one point had 6to4 tunneling enabled by default.
Native IPv6 traffic was stable between 0.1 and 0.2 percent of all traffic, apparently keeping up with the IPv4 traffic growth, but not exceeding it. That's 8 to 16 gigabits. So native IPv6 traffic isn't declining, but it's not increasing either. At this point, we may want to remember Metcalfe's law, which states that the value of a network is the number of users squared. If we replace "value" with "traffic," then the corollary of Metcalfe's law is that the number of users is the square root of the traffic.
So to account for 0.15 percent of all traffic, the number of users who have IPv6 and prefer it over IPv4 must be the square root of 0.0015, which is 3.9 percent. (Only traffic between those 3.9 percent and other members of the 3.9 percent is IPv4 = 0.039 squared = 0.0015. All other traffic has at least one participant that is limited to, or prefers IPv4.) And remember, that's with few Web destinations being reachable over IPv6, few home gateways being IPv6-capable, and few ISPs offering IPv6 to their customers.
One reason IPv6 keeps being a rounding error in the traffic volume numbers is probably that popular Web destinations are reluctant to advertise their IPv6-readiness in the DNS, for fear of users with broken IPv6 setups being unable to reach them. For instance, Facebook and Google both represent a huge chunk of all Web traffic and have IPv6-capable servers, but don't have IPv6 addresses for www.facebook.com and www.google.com in the DNS, so they don't generate much, if any, IPv6 traffic.
Seven weeks from now, on World IPv6 Day, many of those sites will participate in a big experiment to see how bad all of this really is. On June 8 at midnight UTC, participating Web sites will put an IPv6 address in the DNS, so every system that has native IPv6 or a non-automatic tunnel will try to reach those destinations over IPv6. And then we'll really see how many systems that think they have IPv6 but don't are really out there. If you don't want to wait that long, go to test-ipv6.com to see what your IPv6 readiness is.
Update: the text earlier said "the corollary of Metcalfe's law is that the number of users is the square of the traffic."