Predicting IPv6 Growth

Upon hearing the news that Google’s measurement of IPv6 users hitting their websites hit 3% of total users, having just surpassed 2% in September, 2013, I became less skeptical of the exponential growth predictions for IPv6. Under the assumption that people around the world use google and it is atop Alexa’s list of top websites, it would seem such a measurement provides data that could be loosely projected to the Internet at large. To explore this uptick in hits, I sampled some data points from Google’s statistics site in an attempt to create a future projection.

Figure 1: A pair of curve fittings for Google IPv6 users data

In my first attempt at “curve fitting,” I considered quarterly data points going back to early 2009 when google started measuring IPv6 visitors. Applying curve fitting to these data points, I created the the chart of Figure 1, with sample data points represented as blue diamonds. Applying an exponential curve to this data set yields the more gradually sloping (red) curve in Figure 1. As you can see, this curve seems to overcompensate in early years while flattening growth in the later years. This curve’s estimated penetration yielded less than 2% at the end of 2013, well below measured data. I then applied a second-order polynomial trendline in attempt to more closely map to data samples, shown as the steeper (green) line. This curve seems to better fit the later measurements in the chart, though it predicts passing 3% in mid-2014, a metric that has already been reached. Incidentally, the polynomial curve yields an R² of 0.95, which is a better fit than the 0.9 value for the exponential curve.

Figure 2: Curve fittings for most recent Google IPv6 users data

I then considered the data as comprising two segments, the first being the nearly linear component of near zero penetration up through 2011, and the second representing the present growth phase. Considering only these latter data points, Figure 2 illustrates another pair of trend lines, exponential, in this case the steeper of the two and polynomial. Interestingly, the exponential curve yields an R² of over 0.98, representing quite a good fit, and the polynomial fit isn’t too bad at about 0.96. It’s interesting to extrapolate these curves out a couple years to predict the growth in IPv6 Google users and by loose association, Internet IPv6 users. The table below summarizes the predicted percentage of Google users accessing via IPv6, considering data samples back to 2009. These predictions are rather unimpressive, with modest IPv6 penetration with neither reaching double digit penetration even by the end of 2015. But I personally tend to put more stock in the more recent sample data, shown in the far right two columns of the table, illustrating a more rapid adoption of IPv6.

Model prediction	Samples 2009-2013		Samples 2012-2013
End of year	Exponential	Polynomial	Exponential	Polynomial
2013	1.7%	2.3%	2.5%	2.5%
2014	3.0%	3.7%	5.9%	4.9%
2015	5.3%	5.5%	13.8%	8.2%
2016	9.1%	7.6%	32.3%	12.5%

What will the future hold for IPv6 user growth? These models vary quite widely in predicting growth rate, but the trend is definitely upward. Make preparations for IPv6 in your network today.