ARIN and LACNIC Run Rates, End of 2013

It's the last day of 2013, and a good time to consider where we are with IPv4, and where we're going. Geoff Huston shows excellent charts of RIR run rates, and his predictions of runout are the standard for comparison. He includes some additional allocations from IANA in his calculations, and shows a probability of runout for each given week. It's really much better work than my spreadsheets.

I still update my spreadsheets, though, because I like to have my own view, and there's nothing like getting into the data yourself to become more aware of the greater context it provides. I explained how you can create your own spreadsheet in an ARIN guest blog, and I've provided today's version in the Original Documents section here.

For context, let's start with the long view. Here's ARIN runout based on all allocations and assignments since Y2K:

As you can see, it's not a straight line. The rate of allocation has changed several times over the years. Probably the most significant change was triggered by ARIN's policy change at IANA runout, when they stopped allocating based on two years' need, and only assigned three months' need. So let's zoom in to that time frame:

I should explain about the extrapolations. When you use polynomial functions, a flattening curve can make it look like the supply is going to increase. Since that seems unlikely in the extreme, I ignore it; I've left the curves in for the sake of consistency.

In the chart since IANA runout, you can compare the projections and decide which one you think most accurately reflects history, and use that to predict the future. Bear in mind that at APNIC and RIPE, the allocations did increase near the end of their supply. There may be another couple of inflection points in this chart, where trends changed. So let's zoom in again, and just look at allocations/assignments in 2013 (excluding today, 31 Dec 2013):

In the 2013 chart, it's even easier to see how steppy the line is. Every few months there's a significant allocation, and the whole line drops a few points, before continuing at its previous rate. The higher-order polynomial projections (4th order and 6th order) tend to magnify more recent events, so that significant allocation in late November makes it look like ARIN will run out in just 2-4 months. The linear, 2nd and 3rd order polynomials all suggest 9-11 months.

Barring additional IANA allocations, I predict ARIN will run out of IPv4 addresses in Q4 2014.

To resist parochialism, I've done the same analysis with LACNIC run rates. I used to think LACNIC would run out a year or two after ARIN, and AfriNIC a decade after that. Now, we can see that LACNIC is even sooner than ARIN:

In the case of LACNIC, it does seem that the run rate has been accelerating over the course of the year. Because of that, I would tend to believe the higher-order polynomial projections a bit more. That is, the 3rd order and 6th order polynomials are clustered, and more closely track the curve than the linear or 2nd order.

Barring additional IANA allocations, I predict LACNIC will run out of IPv4 addresses in April 2014.

If you're going to make a New Year's Resolution, freeing yourself of the need for IPv4 would be a good one.

Lee Howard
31 December 2013