Users—as in both corporate and consumer—get IP addresses from ISPs. Small ISPs get them from large ISPs, but the big ones get them from the RIRs. The RIRs are:
- AfriNIC—Africa Region
- APNIC—Asia/Pacific Region
- ARIN—North America Region
- LACNIC—Latin America and some Caribbean Islands
- RIPE NCC—Europe, the Middle East, and Central Asia
There are several models to measure depletion of IP address resources. I borrowed the following table from The IPv4 Depletion Site:
| Huston | Hain | Lagerholm | |
|---|---|---|---|
| Freshness | Daily Sat Jan 22 09:40:01 2011 | 2005 with updates from 2008-05-27 | Daily Sat Jan 22 09:40:01 2011 |
| Mathematical model | 2nd order polynomial | Order N polynomials | Exponential and linear |
| Granularity | Sum of RIR | IANA pool | Individual per RIR |
| Fitting method | Least square fit | Least square fit | Least square fit |
| RIR Pool estimates | Fixed low threshold model | No discussion | Fixed low threshold model |
| Smoothing of data | 3 pass with 3 month sliding window | No | No |
| Historic data used | 1200 days (3.29 years) | 2000-01-01 to current (9 years) | 1460 days (4 years) |
| IANA pool depleted | 2011-01-20 | 2010-10-01 | 2011-01-22 |
| First RIR pool depleted | 2011-10-12 | No estimate | 2011-10-09 |
| Last RIR pool depleted | No estimate | 2011-11-01 | 2012-07-28 |
Note the Hain model is not especially fresh, but all 3 show the IANA pool depleted by today (1/22/2011). Of course, models aren't reality.
The IANA Has a table of all the 256 class A (also known as /8) IPv4 addresses. This table shows 7 of these blocks as "UNALLOCATED": 39, 102, 103, 104, 106, 179 and 185 and the page says that it was last updated on "2011-01-04".
Why do the models (and a lot of chatter on network operator mailing lists) say that we're at the end, or nearly so, when IANA says there are 7 /8s available? It's because of a special policy from ICANN governing the coming end days of IPv4 allocation: When the number of available /8s is equal to the number of RIRs, the normal allocation policy goes away and IANA will allocate the remaining /8s, one to each RIR. In other words, once there are 5 left, there are none.
The reason all the chatter indicates that two more have been taken is that APNIC's address pool has reached a point where they are entitled to take two. 2 in this case means 3. But then the IPv4 gravy train runs off the rails.
The next step in the depletion of the pool is for the RIRs themselves to run out. The Huston and Lagerholm models are very close in this regard, both with dates this October, less than 9 months away. At that point, ISPs in whatever region runs out first will no longer be able to get new addresses. Lagerholm predicts the last RIR running out next July. When will ISPs themselves run out? This is even harder to predict.
What happens then? Nobody really knows. There's IPv6 of course, and everyone knows that moving to it is an eventual necessity, but users have been resisting it. A market for the reallocation of IPv4 addresses makes sense economically, but there's no technical mechanism for doing so.
Actual announcements of the IPv4 depletion have not happened, but look for more news on this subject this week, first as the final allocation to APNIC becomes official and then the invocation of the special final allocation policy by IANA. This is big news and it will affect all of us before it's over.
Many thanks to Leo Vegoda of ICANN for help with this article.
Originally posted to the PCMag.com security blog, Security Watch.