You are here: Home > Publications > IPv6 Info Centre > About IPv6 > IPv4 Exhaustion > Archive > IPv4 Address Space Allocation Rate: January 2007

IPv4 Address Space Allocation Rate: January 2007

Three years have passed since we last visited IPv4 Address Space Allocation and much has happend.

Geoff Huston, Chief Scientist of APNIC, has created a web site that keeps track of statistics about address space allocations and address space usage on the public Internet. The IPv4 part of it can be found at: http://www.potaroo.net/tools/ipv4/

On this web site, Geoff also tries to answer the question:

“What is the anticipated date when the current policy regime concerning the distribution of IPv4 address is no longer relevant?”1

Or, in other words, we are looking for some indicators as to the time when our current policies for IPv4 address distribution are expected to run out because the unallocated address pool on which these policies are based is exhausted.

We will not repeat Geoff's statistics collection and modeling work here. Please look at that for an extensive treatment of the subject and form your own opinion.

However we want to point out that the IPv4 address space allocation rate over the past three years (2004-2006) is significantly different from the previous four years (2000-2003). The graph below illustrates this:

[IP Count 200-2006]

The dots represent the amount of IPv4 address space that has been allocated by the RIRs at each point in time. The green dots are taken directly from the allocation statistics published by the RIRs on 31 December 2006 2. The red dots are a slightly more accurate representation; see the Insert for an explanation of this.

Allocations during the years 2000-2003 can be described extremely well by a linear fit. Then, at the beginning of 2004, the allocation rate increased significantly. At this time, almost three year's worth of additional data can also be described well with a linear fit; we see no strong reason to use a different type of function to describe this data. We have also found no systematical error with the underlying data, so we assume that this increase in allocation rate is indeed happening.

[IP Count 1855-2005]

The above graph shows the same data and linear fits over a larger time frame. Note that this independent analysis compares well with the current linear fit in figure 19b of Geoff Huston's report (as of January 2007).

Since address space allocation rates are largely determined by the demand for address space and the allocation policies of the RIRs, we assume that either or both of these are responsible for this change. We have not investigated this further.

This is not a prediction that the RIR address pools will be empty by the time the projection lines cross the ceiling of the available IPv4 unicast address space. Making such a projection is more involved and will in any case be highly inaccurate because neither address space demand nor distribution policies can be accurately predicted. Therefore we leave it to the reader to draw conclusions using their own judgment. For a more detailed treatment of such predictions we refer you to Geoff's report. at: http://www.potaroo.net/tools/ipv4/.

Research: René Wilhelm
Text: Daniel Karrenberg


Insert

We investigated possible systematic errors that could have caused the change in allocation rates. We found that the amount of allocated address at a particular date in the past is over-stated in the RIR statistics files. The reason lies in the way allocations are dated.

Some RIRs reserve a block of contiguous address space when making certain allocations. If the original requester receives another allocation within a certain time, the RIR allocates this reserved and contiguous block. This way a single address prefix can be used for both blocks in the routing system. This is called route aggregation and saves routing table space.

Some RIRs will register the second allocation in the statistics file by merging it with the first allocation while keeping the original date, e.g. the allocation date of the first allocation. This results in an overstatement of the amount of allocated address space at the earlier date. The green dots represent this data including the overstatement.

In order to investigate the extent of this error we have processed all stats files since 26 November 2003, computed the amount of address space allocated at the time each file was created and plotted these in red. This shows that the current stats file typically overstates the amount of historically allocated address space. However it can be clearly seen from the graph that this overstatement does not change the results in a significant way.

References

1.Geoff Huston, IPv4 Address Report

2. Allocation and assignment statistics are published on the RIR ftp sites on a daily basis. Since 26 November 2003, all RIRs publish the data in the same format. Because every RIR mirrors the other RIRs statistics collections, users only have to contact one ftp server to access all RIR statistics. For this analysis we used the data stored under ftp://ftp.ripe.net/pub/stats