[AusNOG] IPv4 Exhaustion date changed to December.

[Geoff Huston]

On 22/06/2010, at 7:48 AM, James Paussa wrote:

> 
> 
> But I also think the article is a little misleading, what I took from it
> was the RIRs are being allocated /8s at a pretty good rate.

The RIRs are allocated addresses according to a procedure described in http://aso.icann.org/wp-content/uploads/2009/09/aso-001-2.pdf

The current rate of address consumption is some 200 million IPv4 addresses per year, with seasonal variations of +/- 16% or so.


> From my
> understanding the RIRs have to allocate IP addresses in their region then
> they get used by ISPs etc. Now, IANA may run out of allocatable IP
> addresses,


currently, assuming a continuation of the current dynamics without undue shocks, this will occur mid next year, or some 13 months from now, but the assumption of smooth continuity into the future may not be a good one. At that point in time IANA will allocate to each RIR its "last /8" and at that point for IANA the current allocation policy framework for IPv4 address allocation will come to an end.


> then the RIRs have to run out,

this will take a variable amount of time. Assuming that the RIRs are able to efficiently use the remaining free space in the old B and C address blocks and assuming that the demand model is a smooth continuation from today (ho ho!) then the first RIR to exhaust its own pool and come down to its "lazst /8" will be APNIC, in April 2012. Frankly thats assuming a lot about human behaviour in the face of looming scarcity.


> then the ISPs have to run out of
> their stock.
> In an end of the world type scenario I think we are looking for D Day to
> be in a couple of years time. We will start by charging for IPs to end
> customers at an inflated rate to reduce their use, then conserving them
> through SP NAT (look at 3G networks now) and finally adopting IPv6 when
> there is no other choice.


A typical market reaction to scarcity is the imposition of a "scarcity rental", where access to the increasingly scarce resource attracts a premium price. However, this situation is not quite so clear as is portrayed here, as the issue here is the imposition of further costs on the existing IPv4 user base of clients and servers and service providers to go from IPv4 to IPv4 / Ipv6 dual stack, buth without immediately realisible immediate benefit. The economics of such a transition are by no means clear, even at this point. (http://www.potaroo.net/presentations/2010-06-11-ipv6-transition.pdf has some further thoughts as to why this transition presents some really quite unique aspects and challenges. )

regards,

   Geoff

   (speaking for myself here, and not necessarily for APNIC)