After reflecting on the data from IPv6 day, the phrase the best comes to mind is: “Wagon’s Ho!” It’s going to be a long hard slog to IPv6-Land. Yesterday’s IPv6 flag day looks to have been a success. After a decade of implementation work by the infrastructure vendors in building towards IPv4 functional parity, combined with the months of preparation by the content and service providers in constructing the routing and namespace frameworks, IPv6 day came off successfully. For a 24 hour period starting at midnight UTC, anyone with IPv6 access connectivity could get to some of the largest content providers’ data through their v6 stacks. With six of our customers’ help, we were able to get a glimpse into some of the details of the day.
Figure 1. Application breakdown for native IPv6 traffic from our six carrier partners.
For a bright 24 hour period, shown in Figure 1, the IPv6 network looked a little bit more like its IPv4 big brother. As we have shown in some of our previous posts, the traffic mix for the IPv6 network could be best described as flotsam and jetsam: encrypted file transfers, peer to peer traffic, and experimental protocols. However, during yesterday’s v6 day the mix was dominated with web traffic. The proportion of web traffic grew during the day up until the midnight cutoff point where some of the major content providers withdrew their namespace support. At midnight UTC the web traffic falls of the cliff and the traffic mix returns to its pre-v6-day chatter.
Figure 2. Percentage of IPv6 traffic of all Internet traffic in our six carrier partners.
Figure 3. Proportion of Native IPv6 traffic versus total IPv6 traffic in our six carrier partners.
The good news shown in Figure 2 is that IPv6 traffic roughly doubled during the v6-day period. However, doubling a fraction of a percent, is still a fraction of a percent. These datasets correlate anecdotally with those from a lot of providers that we’ve talked with at our customer summit this week in Amsterdam. Combined with the results shown in Figure 3, they lead us to think more about the access side of the Internet. Even with a quad-A record in their DNS response, we just didn’t see Internet clients switching over in mass to the IPv6 content servers. It’s not surprising when you think about the level of indirection almost all subscriber side internet connections use. Unless you plug your home PC directly into the cable/dsl modem (another potential point for v4/v6 breakage), you probably have at least a mediating DNS caching device (home router or wireless basestation) that may not elegantly switch back and forth from v4 to v6. The inertia and complexity of changing this element of the Internet is massive.
Again, I’m left thinking of a Wagon train metaphor. Yesterday was the start of a long road that will take us to IPv6-land. It will pass through the more efficient use of IPv4 address space due to market forces; the operational pressures to not run two separate networks that need debugging and maintenance; and the not insignificant security threats introduced by the incredible complexity and new boundary conditions of the juncture of v4 and v6 networks. I have no doubt that the wagons will eventually get there, but it will not be easy.
Additional resources:
Internet Society: Participant Dashboard
RIPE NCC: World IPv6 Day Measurements
Google: World IPv6 Day
Akamai: IPv6 Statistics
Comcast: IPv6 Day Q&A
Comcast Leverages Arbor Peakflow SP to Facilitate IPv6 Transition: Featuring John Brzozowski, Comcast’s Chief Architect for IPv6
Leave a reply
