Lightning Poll: Three Questions on IPv6

A couple of weeks ago, I ran what I called a "Lightning Poll" at NANOG60, in which I asked three questions:

  1. In what year do you think IPv6 will be cheaper than IPv4 or dual-stack?
  2. In what year do you think IPv6 will offer better performance than IPv4?
  3. In what year will a significant number of users have IPv6 only?
I poted the poll during a Lightning Talk, then let people answer it for 5-10 minutes, then took it down and parsed the results.

What I wanted from this poll was a sense from the NANOG community of when IPv6 would tip in the adoption curve. I figured that NANOG attendees (including remote participants) are the best informed community about the current and future deployment of IPv6, since they are the ones who run networks.

Here are the results:

Question#ResponsesAverage80th pctile
In what year do you think IPv6 will be cheaper than IPv4 or dual-stack? 111 2017 2018
In what year do you think IPv6 will offer better performance than IPv4? 110 2016 2019
In what year will a significant number of users have IPv6 only? 109 2020+ 2020+

Breakdown by year:

Question2014201520162017201820192020 or later
Cheaper1310181130613
Faster431218610510
IPv6-only021012211114

How realistic are these dates? I wondered whether the NANOG audience would be friendly or hostile, or firmly realistic about the prospects for IPv6. Compare this survey to the slightly smaller one I ran on in an IPv6 group on Facebook last November, and the NANOG community is definitely more conservative than the IPv6 community. One would expect an IPv6 community to predict earlier (more bias-confirming) dates than a neutral NANOG, and this does seem to be the case.

In answering the first question, when IPv6 will be cheaper, 50% of respondents answered 2017 or sooner (compared to 89% in the IPv6 Facebook poll). An additional 30% said the following year, 2018. That seems like a fairly strong consensus prediction that in 3-4 years, IPv6 will be cheaper. Since the question specifically includes dual-stack, that suggests that by 2017 or 2018, it will be more expensive to run IPv4 than not to run it.

The answer distribution for the question about when IPv6 will be faster is interesting, since I had just shown how much faster IPv6 is. It should not be surprising then than 42% of respondents said IPv6 will be faster in 2014 and over 50% by 2015. What surprised me was that 70% said by 2016, and it took until 2018 for over 80% to predict a faster IPv6. Still, with good data from multiple sources with differing methodologies, reinforced by the median, I stand by my statement that IPv6 is already faster than IPv4.

Maybe the most dramatic prediction is when IPv6-only will represent a significant number of users. I could have included a span of 30 years, and people would have used all of them, because I think people read this question as, "When will IPv4 be turned off." Those are, of course, different questions. Even so, a cumulative 64% said there would be a significant number of users on IPv6-only by 2018, just four years from now. 80% said 2019.

At the risk of reading too much into these results, I conclude:

  • IPv6-only will be cheaper than dual-stack or IPv4 only by 2017, maybe 2018.
  • IPv6 is already faster than IPv4.
  • There will be a significant number of IPv6-only users by 2018 or 2019.
  • Therefore, there are good reasons to support IPv6 by 2017, but it's absolutely critical for your business by 2018.

Call to Action: IPv6 by 2017

Based on these results, plus everything else I've said everywhere, I propose that 2017 be used as the target year for complete IPv6 deployment.

Some places may need the extra year, to 2018. However, to reduce costs, you need to be complete by 2017. And if there are a significant number of users without IPv4 by 2018, then there will be some few by 2017. Finally, there are already performance benefits to running IPv6.

Lee Howard
5 March 2014
@WLeeCoyote