28

Look at the graph of IPv6 adoption rates maintained by Google here:

https://www.google.com/intl/en/ipv6/statistics.html

Zoom in to the September to December 2015 period.

The graph of IPv6 adoption rates is clearly periodic, with much higher rates at the weekend. Why is that?

Peter Mortensen
  • 2,318
  • 5
  • 23
  • 24
DavidA
  • 383
  • 3
  • 6

2 Answers2

32

tl;dr: Because Comcast.

Comcast has, by any measure, the largest IPv6 deployment in the world, with the greatest number of users. Commercial/business networks are lagging behind with regards to IPv6. People are not at work on the weekends and such we see higher IPv6 adoption then. I'm sure other residential ISPs contribute to this trend as well.

Say what you want about Comcast's business practices. No one can fault them for being a staunch IPv6 advocate from very early on. (of course they were forced into it due to IPv4 not having enough addresses for them to manage their own device pool)

EEAA
  • 109,363
  • 18
  • 175
  • 245
  • Makes sense, everyone turns on and connects their bazillion IoT devices to their home wifi. – Colyn1337 Dec 29 '15 at 14:19
  • The trend started a lot earlier though. Even back in the earliest days of those statistics (when 6to4 was twice as common as native IPv6), the trend was visible on both 6to4 and native IPv6. – kasperd Dec 29 '15 at 14:51
  • @kasperd Perhaps another additive effect is that Youtube is available over v6 and people youtube more on weekends/holidays? – EEAA Dec 29 '15 at 14:58
  • @kasperd Windows Vista and 7 shipped with 6to4 enabled, (which in retrospect was not such a great idea) so they would automatically attempt to connect to public 6to4 relays and make outbound IPv6 connections via the 6to4 tunnel. But the fact that IPv4 is still preferable to 6to4 kept traffic low. – Michael Hampton Dec 29 '15 at 15:27
  • @MichaelHampton I think Apple routers were first with that. It may look like a bad idea in retrospect, but it was part of the reason we got RFC 6555, which is a good thing. And I believe RFC 1918 has done much more harm than 6to4 has. Some people have been fighting 6to4 so fiercely that they forgot why they were fighting 6to4 in the first place. For a few months this even caused a decline in native IPv6 adoption. What really sucks is the fact that lack of native IPv6 deployments means that for some users 6to4 and Teredo are actually the most viable strategies. – kasperd Dec 29 '15 at 15:45
  • @kasperd Well, [public 6to4 has been put out to pasture](https://tools.ietf.org/html/rfc7526). Teredo is sufficiently nonfunctional for many people that it's probably not far behind. – Michael Hampton Dec 29 '15 at 15:57
  • @MichaelHampton My experience with Teredo is that if you spend 5 minutes configuring it correctly, then Teredo will be more reliable than running TCP through a CGN. – kasperd Dec 29 '15 at 16:01
  • @kasperd Yeah, but Joe Sixpack isn't going to manage to configure Teredo by himself. – Michael Hampton Dec 29 '15 at 16:13
  • @MichaelHampton A lot of ISPs have assumed for the last 17 years that it made sense to deploy IPv6 at the rate at which end-users understood how it works and asked for it. That has not exactly gotten us very far. At the moment server administrators are better off assuming that between their servers and the end-users, there will be an unreliable ISP. And you are better off doing any mitigations you can server side. Having your own relays for outgoing traffic to users with 6to4 and Teredo is part of that. The deployment of such relays are not deprecated by RFC 7526. – kasperd Dec 29 '15 at 16:26
  • This doesn't really answer the question though. It answers who, but not why or how. Are you saying that simply a large number of comcast users only log onto the internet on the weekends? Or is it more complex like comcast is slowly pushing large batches to IPv6 on weekends? – cde Dec 30 '15 at 05:23
  • @cde Comcast users are at home on weekends. Hence higher usage. – EEAA Dec 30 '15 at 05:23
  • 2
    @EEAA every chart I find online shows higher usage during the week than on the weekend. See most of the results here https://www.google.com/search?num=30&safe=off&client=safari&rls=en&q=internet+usage+by+weekday&oq=internet+usage+by+weekday&gs_l=serp.3..0i22i30.71756.80284.0.80635.33.19.8.6.6.0.132.1966.7j12.19.0....0...1c.1.64.serp..8.25.1599.vCsVLyHWhGc – cde Dec 30 '15 at 05:43
  • 1
    Even better, this quora with links and graphs to Quantcast traffic tracking. Most damning, even Google Search has much less traffic during the weekend compared to the week (Desktop, not mobile) according to their live event and Amit Singhal https://www.quora.com/Does-website-blog-traffic-usually-go-up-or-down-on-weekends – cde Dec 30 '15 at 05:49
  • @cde The charts you are referring to show absolute traffic, Google's IPv6 chart shows IPv6 traffic as _percentage relative to total traffic._ You could argue, though, that the lower traffic from (IPv4-only) company networks pushes the ratio of IPv6-enabled home users higher. – Dubu Jan 03 '16 at 16:30
5

Some mobile networks have already switched a large portion of their customer base over to IPv6 (T-Mobile being one that comes to mind).

If we make the plausible assumption that people tend to spend more time away from their home computers on weekends, then it makes sense that IPv6 traffic would increase proportionally.

David
  • 715
  • 1
  • 5
  • 6
  • 1
    I don't think this explains that much of it. Android devices are pretty much constantly in contact with Google, for instance, and so they would be counted every day of the week. – Michael Hampton Dec 30 '15 at 03:00
  • @MichaelHampton - but they're also *likely* on wifi most of the time during the week, whereas you're far more likely to be away from wifi on the weekends – warren Jan 04 '16 at 21:40
  • @warren I don't know about you, but many people are not on wifi while commuting to and from work! – Michael Hampton Jan 04 '16 at 21:44
  • @MichaelHampton - and while they're commuting (if they're driving), they're not using their phones (and if they're on subways, they can't get reception anyway). And I'm going to bet while at work, they're on wifi (just like they'd be at home), both of which are substantially longer than the amount of time they're just on cell data – warren Jan 05 '16 at 03:11