The decline in IPv4 addresses continues (and the IPv6 debate)
by Andy Oram
Tired of endless news reports about Super Bowl football? Take a moment
to view some really hot stats! The recent 2006 IPv4 Address
, besides featuring a nice cover of O'Reilly's BGP book,
details the trend in giving out IP addresses.
The bottom line is that the world has used up about two-thirds of the
available IPv4 addresses, with about 1.3 billion addresses left. This
suggests that we have not hit a crisis yet and are still not impelled
to adopt IPv6. But there are some interesting details.
Actually, Asia has been feeling the IPv4 address crunch for quite sometime now. Remember that tussle a while back about the entire country of Qatar being blocked by Wikipedia? That was because the entire country of Qatar was being provided Internet access via a single IP address (which was responsible for some wikipedia spamming). Entire countries being served by 1 NAT address is not that common yet but entire large networks are.
In the Philippines, the norm is to only get a /24 address block from the Internet Service Provider even if your organization has a couple of hundreds computers. Using NAT is the norm in these parts of the work. All our mobile phone operators do not assign globally routable address to mobile phones and mobile broadband users.
This is reality of the shortage. So far, most organizations are managing. But, for how long?
The anecdote about Qatar's "single IP address" is grossly misleading; the truth is quite illuminating with respect to where the bottlenecks really have been to the present.
There's no denying the finitude of IPv4 addresses, and the inevitability of exhaustion over the next decade or so, but at present the perceptions of "crunch" are only that -- perceptions. IPv4 addresses are still available from the RIRs to independent service providers in qualtities that are limited ONLY by the degree to which a credible requirement can be demonstrated -- "credible" meaning "to enable real, not vapor services, once other necessary inputs (staff, equipment, network links, etc.) are in place." Qatari institutions currently enjoy exclusive beneficial control of tens of thousands of public IPv4 addresses, and most of these are actively originated to the rest of the Internet by Qatar Telecom or the Qatar Foundation. The fact that many individual Qatari users were critically dependent on a single proxy server / IP address speaks to design (or policy) choices inside Qatar, and NOT to the inability of ISPs there to secure public IPv4 addresses.
Similarly, to date most mobile phone data networks have been explicitly designed to keep users segregated within a carrier-controlled service domain. It's not the lack of public IP addresses, but the unwillingness to facilitate 3rd party service competition, that generally limits the functionality of such networks.
The "supply chain" for public IPv4 addresses begins with self-governing regional institutions that devote a great deal of time and effort to admnistering a "fair" and "efficient" distribution process; they try hard to balance the measureable Internet production requirements of the present against the foreseeable Internet production needs of the future. However, in many cases the IP address supply chain will include other "redistributors" -- perhaps a national-level registry, perhaps one or more commercial service providers -- which may define "fairness" and "efficiency" differently, and which may be considerably less responsive to constituency expectations and feedback than the RIRs at the top. More often than not, concerns about (un)fairness and (un)availability can be traced to one of these intermediate suppliers. In general, such cases reveal nothing at all about the overall status of IPv4 addresses and the IPv4 address distribution system -- except of course that individual ISPs can and do exercise independent judgment in how they manage addressing for their own customers and their own service domains.
- As the article mentions server consolidation is taking place, but this requires huge ip address space. In our environment is not uncommom to have 10-20 address per server boxes. And we are planning to have 50-100 addresses per server boxes.
- Your argument about the immediate alarm is false. Currently systems are engineered with NAT in mind, because of scarsity of ip addresses. This make system less scalabel, more complex to operate and more likely have bug in their codes.
Quick reponse to Janos and the interested general readership... the attachement of (sometimes quite) large numbers of IP addresses to individual devices is commonplace, and of course well-understood by all of the institutions that play a role in the distribution of IP address resources. Many devices have multiple addressable interfaces or independent processes (e.g., access servers), and as colo space continues to command an ever-growing premium, more and more such devices are being squeezed into ever-smaller increments of space. While some *transition plans* between one kind of hardware platform and another might require a large numbers of addresses on a temporary basis, there are usually ways of working around this ... it's not clear why something other than absolute growth in demand (e.g., for Internet access, content, services, etc.) should increase the absolute demand for public IP address resources permanently.
As for the need for and consequences (reduced scalability, flexibility, increased complexity) of NAT, this has been a fact of life for over a decade -- it's not good (!), but neither is it cause for exponentially greater hysteria today than it was in the 1990s. It will be a great day when (if) the finitude of public IP addresses ceases to be a design consideration, maybe... but scarcity is not the only demand driver for NAT, and the scalability, complexity, buginess, and aggregation requirements of whatever replaces NAT will no doubt help to keep network engineers complaining (and fully employed) for many years to come ;-)