Forecasting when IPv4 addresses will run out

There have been various viewpoints and forecasts as to when the available global IPv4 addresses will run out.  Among them, Geoff Huston from APNIC, in July 2003, said that the rate at which addresses are being assigned is almost linear and predicted that IANA’s pool of unassigned addresses would be exhausted in 2021, RIR’s in 2022.  Since he used RIR’s detailed and reliable data for his analysis, his prediction has come to be used widely since then. (see endnote 1)

However, after this prediction was announced, the number of IPv4 addresses being allocated started demonstrating an upward trend.  According to the totaled results of RIR’s public statistical data, the number of allocated hosts from January to October 2003 was about 66.64 million hosts (3.97x/8).  For the same time in 2004, it was 111.04 million hosts (6.62x/8); for 2005, it was 148.54 million hosts (8.85x/8).

Seeing this upward trend, people started reconsidering the projection of when IPv4 addresses will run out around summer 2005.  First, Tony Hain from Cisco announced, in the September 2005 issue of Internet Protocol Journal, that “if the current allocation rate continues, the IANA’s unallocated address pool will run out in 2009 at the earliest and 2016 at the latest”.  (see Figure 1)

Figure 1: IPv4 Address Lifetime Projections by Tony Hain

/8 number

So, Geoff Huston also reworked his prediction on when IPv4 addresses will run out.  As of November 9, 2005, he predicts that IANA’s unallocated address pool will be exhausted on May 12, 2012, and that RIR’s will be exhausted on May 27, 2013. (see endnote 2)  Both predictions show that their exhaustion times have been moved up considerably from the prediction made in 2003.  Since Geoff Huston’s prediction is about halfway between Tony Hain’s predictions, it is probably fair to say that sometime around 2012-2013 is currently the most probable time for the IPv4 address pool at RIR to become exhausted.

The reason the projected time has gotten sooner is because, as I mentioned above, the rate of IPv4 address allocation has increased recently.  Then, why is the allocation increasing now?  I’ll list the possible reasons below.

1. As you can see in the United States and China, as well as in Japan and Korea, -always-on Internet connection/broadband connections within households have spread internationally.  So, compared to the time when people were connecting to the Internet via dial-up, the number of necessary addresses per user has increased tremendously.
2. There are now more applications, such as IP telephones, online games, etc., that require users to have a global IP address in order to receive services.  Because of this, companies that provided private IP address assignments up until now are switching to global address assignment.
3. The enforcement of RIR/NIR address policies is not as strict as it used to be.  Among the changes are the following: they now let users decide whether they adopt NAT or not at their own discretion; and there are cases that global addresses are assigned even to private networks that exceed /8.  Also, a system called “assignment window”, where the assignment is done at the companies’ discretion, has started functioning well, and the assignment process has gotten smoother because the applicant now faces less obstacles.  So, you can’t ignore the fact that the allocation number is increasing as a result.

Of course, the data I’ve shown above is only a prediction, and the actual year that addresses run out will depend on various conditions.  However, when considering that situations like 1-3 above will continue to happen all over the world, you may think that the IPv4 allocation rate will speed up even more, but it is hard to think that it will actually slow down.  In addition, I have no doubt that there will be a last-minute rise in demand when people try to apply for addresses before they run out.

On the other hand, factors that can delay when addresses will run out also exist.  The IPv6 deployment state is one of the big factors.  Geoff Huston suggests the recovery/reuse of unused addresses as a factor.  However, Tony Hain doesn’t think that the return/reuse of unused addresses are being done efficiently.  So, it is appropriate to think that the recovery/reuse of addresses will only allow IPv4 addresses to last a little longer.

What will IPv4 address allocations/assignments be like when people can see that addresses will run out within 2 years?  It doesn’t mean that the networks of companies who have already been assigned IPv4 addresses won’t work anymore.  However, on the other hand, ISPs have no choice but to stop accepting new IPv4 customers.  In such circumstances, quite a bit of commotion around the issue of address allocation/assignment can be expected, including the above-mentioned last-minute rise in demand, black-market address sales depending on the situation, complaints about unfair address distribution among countries and organizations, etc.  I don’t think they can continue to assign addresses until the very last minute of its expected lifetime.

The fundamental, and only, response to this is probably IPv6 adoption and deployment.  For example, in the TV broadcasting field in Japan, it has already been decided that they will stop analog terrestrial broadcasting service in 2011, and various measures towards deployment have been taken at the government and private level.  For Internet-related matters, I must say that the move to take action and adopt IPv6 has been progressing slowly compared to that, particularly considering the fact that IPv4 addresses may run out even before analog terrestrial waves are stopped. 

IPv6 products and technologies are already becoming ready, and it is necessary to promote IPv6 adoption in actual systems.  It will probably be necessary to do things such as building new systems using IPv6 and making systems that need updates IPv6-ready.  By the time the next upgrade is necessary, it is possible that IPv4 addresses will have run out.  Considering the possibility of needing an IPv6 deployment between updates and its cost demerits, you could say that there are enough reasons to make the system IPv6-ready from the beginning.

In this article, I have given the latest news on the rate of IPv4 address allocation and explained its impact.  With regards to Tony Hain’s 2009-2016 prediction, we will need to continuously monitor the allocated amount and analyze it in order to see how correct his prediction will be.  We, at JPNIC, are also planning to start a project to do this on our own.  I would like to report on it accordingly.

Endnote
1. The ISP Column : IPv4 - How long have we got?
http://www.potaroo.net/ispcol/2003-08/ale.html

2. The predictions are updated every day.
http://www.potaroo.net/tools/ipv4/