Feel free to send questions or suggestions to the email address on the bottom of this page.
This page summarizes the results of a statistical analysis on assignment of IPv4 addresses over the last several years as reported by the American Registry for Internet Numbers. The viewpoint of this analaysis is from the perspective of ARIN who administers blocks of IPv4 addresses to other registries, Internet Service Providers (ISPs), etc. The address/mask pairs are listed along with a date when the block of addresses was directed from ARIN. There's no attempt to determine what proportion of addresses within a block are currently being used by the ISPs (see Ipv4 Address Utilization in Related Inforation): the goal is to determine in what sort of time frame ARIN becomes affected by limitations of the 32-bit address space.
The address dump used includes overlapping address ranges, usually with one address block being a subset of another. This can be interpreted as a second-order delegation of addresses within the larger block or as the larger block having been surrendered by its original owner at some point in the past and the smaller blocks representing the redeployment of the addresses to other owners. Since the dump does not indicate which scenerio affects any given address region, I am assuming the former in all cases. This would have the effect of overestimating the historic values, reducing the overall growth rate towards the current levels.
The raw population numbers were scaled down into percentages of the total available numbering space to make the values easier to relate to. The model, however, does not enforce a hard limit at 100% since what I'm trying to gauge is the total demand for IPv4 address space if the supply were infinite.
One point that must be emphasized is that these are trends and not predictions. Past performance does not guarantee future results, your milage may vary. Before extrapolating these lines out into the future, one must recognize that these curves are estimated by looking at historic data and are in fact only modeling IPv4 address usage within the same areas that they have traditionally been associated. The implicit assumption is that there is not some form of paradigm shift that either dramatically increases or eliminates the demand for new addresses in the near future.One aspect of the current paradigm that bears special consideration are the policies that ISPs are following in reguard to delegating portions of IPv4 address space to their subscribers. If they were to become less concerned about conserving IPv4 address space, there would likely be greater demand in the short-term and the curves would be raised proportionally.
The "Address Space Delegation" graph represents the best fit
against the ARIN's report for address space usage.
As of 30 October 1999, ARIN reports that 43.65% of
the IPv4 address space as having been at least delegated to other
providers.
The estimated curve suggests that the total demand levels
off at just over 45.44% of the address space.
This estimate needs to be qualified, however, since the
model has been underestimating the actual data for
the last several months now.
As more data becomes available over the next few months
and exerts a greater influence upon the rest of the distribution,
the trend line will be pulled upwards.
This "Variance" graph show the 95% confidence interval
on the estimate of the limit that the delegation line approaches.
This is to determine how consistant the estimate
for the size of the resulting population
the `C' parameter of the model
has been as new data becomes available.
This illustrates the qualification in the previous paragraph:
the last point is the confidence interval on the estimate
based on the currently available data (45.44 +/- 0.84%);
the previous points in the graph
are the confidence intervals on the regressions
when I drop the data after the corresponding date.
By way of example, I drop all of the network numbers that are
listed as having been delegated between 1 Octover and 30 October
then run the regressions again to reach the value at the next-to-last
point.
If that is an accurate representation on what the state of the
world was on 30 Sept 1999, the estimate would have been
45.41 (+/-0.86)% of the total address space.
One will note that the estimates for maximium address space utilization have risen about 6.2% over the last 3 years. If one were to argue for extrapolating this rate linerally into the future (not likely supportable, given the slowing growth rate at the end), the resulting statement would be that "in the year 2007, the trend line will suggest that we will eventually run out of IPv4 addresses". Over the last few years, the difference between the current and eventual delegation rates has been relatively small, though, so one might assume that the amount of time left at that point would be relatively short.
So what caused the rate of change to start to drop off? This could have been either a natural effect ('there never was a problem') or could have been a result of ARIN's and the ISP's more stringent address allocation policies ('there was a problem but we saw it in time and addressed it', so to speak).