Network World ran an article I wrote in January of last year discussing the decline in IPv6 address allocations since 2002 and the prediction that the tide would begin to turn. Now that 2007 is behind us, it is worth a look at last year's statistics to see where things stand.
Indeed 2007 was a big year for IPv6. The five regional Internet registries (RIR) -- tasked by the Internet Assigned Numbers Authority to govern IP address allocations -- made a total of 379 IPv6 allocations last year. That is about 70% growth from 2006 and close to the 2002 peak, a dramatic jump from what was a 5-year pattern of decline. Clearly there is interest and movement towards IPv6.
All RIRs made more allocations in 2007 than in 2006 with the exception of the African Network Information Centre (AFRINIC), which was just slightly down. But AFRINIC demonstrated the most growth in its overall allocation base at over 60%. The Latin American and Caribbean Internet Address Registry also continued on a fast pace, with the Asia-Pacific Network Information Centre and the RIPE Network Coordination Centre in Europe moving steadily along as well.
The big story is the American Registry for Internet Numbers (ARIN), whose 114 allocations represent about 42% growth to its allocation base. ARIN serves the United States, which is often said to have little interest in IPv6 and is behind the rest of the world. That appears to be changing, most likely as a result of the self-imposed June 2008 government mandate for IPv6 compliance.
It is worth noting that most allocations made were to service providers. Until 2007, RIR policies dictated that enterprises must acquire IPv6 addresses from their upstream service providers. There were no policies that allowed enterprises to acquire provider independent (PI), or “portable,” allocations directly from an RIR. Although this policy had the good intentions of promoting address aggregation and controlling routing table growth, it fell under scrutiny for various reasons.
First, the policy contradicted the primary motivation for IPv6, which is the dwindling IPv4 address space, the difficulty many organizations have in acquiring addresses and the many side effects that presents. The policy lead to this conundrum: “I can't acquire IPv4 addresses and am being told to migrate to IPv6, but despite IPv6's virtually infinite address space, I still can't acquire my own addresses?” Notwithstanding the benefits of aggregation, this contradiction was tough to defend.
Second, the policy creates an anticompetitive situation. Regardless of IPv6 features that make renumbering easier, changing a live network's address scheme is always logistically complicated and disruptive. An enterprise may decide to stick with a service provider it may be unhappy with simply to avoid the risk and effort involved in renumbering.
Third, the policy created problems in multihoming scenarios. In today's world where application and network availability are truly mission critical, creating redundancy through multihoming is a must for many organizations. In IPv6, unlike in the IPv4 Internet, it is policy that service providers only announce their aggregate blocks and not that of other service providers. This creates a hardship for those who want to have multiple upstream providers.