IPv4: Is the Party Over?

Last week I focused on risk mitigation through reduction of human errors. This week I'd like to talk about another risk, not to your network but to your business. Specifically, the inability beginning sometime in the next four years to expand or add any new services that require new IPv4 address allotments. There won't be any.

Such talk is almost always dismissed as Chicken Littleism. But the numbers don't lie. There are a couple of good studies examining the allocation rate of IPv4 addresses done by APNIC's Geoff Huston and Cisco's Tony Hain. Huston predicts that the IANA's pool of unallocated IPv4 addresses will run dry in mid-2011, whereas Hain predicts somewhere in 2010. What is perhaps more interesting is that Huston began, in 2003, predicting that the IANA pool would deplete in 2021 and Hain's first study predicted 2008. As both have refined the fitting of curves to the data, and as the IANA allocations rates have provided more data, Huston's conclusions are moving much closer while Hain's are moving a little further out, and they are converging to somewhere around late 2010.

But you don't need the statistical projections of either Huston or Hain to tell that things are moving fast. A more casual look at the facts tells you the same. Let's start with some basic background, for those of you who might not already know how IPv4 addresses are allocated.

The 32-bit IPv4 address space can be divvied up into 256 8-bit chunks, or /8 (pronounced "slash eight") blocks. That's all there is, or ever was: 256 /8s. The IANA allocates /8 blocks to the five worldwide Regional Internet Registries (RIRs), which then allocate pieces of the blocks to Local Internet Registries (LIRs) or other large organizations. Each RIR assumes an 18-month supply of assignable addresses, and requests more from the IANA whenever its available pool either falls below 50% or when the pool is no longer sufficient for a further 9 months of allocations. The IANA allocates one or more new /8s to restore the RIR's pool to an 18 month supply. With that overview, let's look at the facts.

At the beginning of 2003, 92 of the 256 total IPv4 /8s remained in the IANA's unassigned pool. That doesn't look too bad.

  - In 2003, 5 /8s were allocated to the RIRs.

  - In 2004, 9 /8s were allocated to the RIRs.

  - In 2005, 11 /8s were allocated.

  - In 2006, 10 were allocated.

So at the beginning of this year, 55 /8s remained. A 40% decrease in 4 years. Already this year, 7 /8s have been allocated to RIRs: 5 to APNIC (Asia), and 2 to RIPE (Europe). ARIN (North America), LACNIC (Latin America), and AfriNIC (Africa) have yet to come to the trough; nor are APNIC and RIPE necessarily finished making requests for the year. It can be assumed that a total of 12 - 15 /8s will be allocated in 2007.

Do the arithmetic however you want, based on these numbers. Here's one view: Given that there were 55 /8s at the beginning of 2007, if 14 of them are allocated every year from now on we have a little less than 4 years left. But the numbers since the beginning of 2003 show that allocation rates are not flat, they are trending upwards.

Those are the numbers. But other factors make the near future of IPv4 even more interesting. I'll let you chew on what I've presented so far, and go into those "other factors" in the next post.