I’ve been thinking about the rate of demand for IPv4 addresses. We can see demand on LACNIC and ARIN bringing them to the late phases of runout policy. Since APNIC and RIPE have entered their stingy final phases, the supply has nearly dried up. With no supply, the number of allocations made does not reflect demand. There’s an alternate supply, in the IPv4 transfer market, so I’ve been wondering about the volume of addresses being transferred.
I’ve looked briefly at ARIN transfers before, and predicting future transfers is a keystone of my supply analysis, which is a significant follow up to my pricing paper, which compares the cost of CGN, buying addresses, and deploying IPv6, not to mention my Game Theory ideas.
Three RIRs publish lists of transfers: ARIN, RIPE, and APNIC (in tediously different formats). Also, ARIN separates out “specified transfers” (i.e., the market), from those due to mergers and acquisitions; RIPE and APNIC consider them a single list. ARIN staff has noted that sometimes organizations use the specified transfer policy rather than the M&A policy, when either could be applied. Therefore, I’ve included both.
When I combine the quantity of addresses transferred over time into a single global number, I get this chart.
This is a cumulative chart; a total of 22.5 million IPv4 addresses have been transferred. Since about February 2012, the demand for IPv4 addresses on the market has been very steady, and pretty linear. That’s significant, and interesting: demand stablized a year after IANA and APNIC runout, not immediately after.
Here’s just data for ARIN:
A subset of those transfers are transfers to organizations in the APNIC region:
Here’s a chart of RIPE transfers:
And APNIC transfers within the region:
Except for the ARIN region, where demand both from the RIR and on the market is very bursty, the volume of transfers over time is pretty consistent. It’s just barely increasing over time.
In Assessing the Supply I looked at NRO statistics prior to runout, and guessed that demand might continue to increase. That turns out not to have happened. Instead, once APNIC and RIPE hit their final austerity policies, demand has held flat in 2012 and 2013, and is on track to remain flat or slightly down in 2014 (LACNIC and ARIN runout notwithstanding).
If demand remains consistent even following LACNIC and ARIN IPv4 free pool exhaustion, and if my analysis of the market in Assessing the Supply is correct, then (following the red line in Figure 3 of that document), then there is only enough IPv4 address supply for three years. That’s based on a naive assumption that demand would not be affected by price, or that price would not change. Alternately, IPv4 prices will rise, which will reduce demand as potential buyers look for alternatives, such as CGN or IPv6.
Interestingly, that’s consistent with the 50% of poll respondents who said that IPv6 would be cheaper than IPv4 by or in 2017. That seems to be a pretty clear date.