[arin-ppml] How much time would we have?
Howard, W. Lee
Lee.Howard at stanleyassociates.com
Thu Aug 28 13:14:54 EDT 2008
At various levels of assignment efficiency (i.e., doing nothing,
marketizing addresses) and at various levels of demand (extrapolation of
current level, increase due to panic, decrease due to IPv6 uptake or
IPv4 efficiency), what is the effective runout date? i.e., can we
various runout dates based on various assumptions?
For each set of assumptions, what is RIR runout date?
Demand assumptions: What multiple of current rate (13.7 per year, plus
Supply assumptions: What percentage of legacy space could be recovered
reallocated (whether by reclamation or by market incentives)?
Demand Assumption (what multiple of current
1/3 Current 3X
0% 2011 2011 2011
30% 2011 2012 2011
50% 2014 2013 2012
70% 2015 2014 2012
90% 2017 2015 2013
(formatting note: I used tabs for whitespace)
In English: for each level of demand (1/3 of current rate, current
rate, or 3X current rate) and each level of legacy space reassigned,
when do we run out of space?
Changing IPv4 policies (harder or easier to get more) doesn't have much
effect on the runout date by itself.
Becoming more efficient in legacy space buys a year or two (or more,
depending on how inefficient you think legacy assignments are).
If you want to delay the time-to-runout by two or more years, you need
extreme measures: significantly reduce the amount of IPv4 address space
assigned by RIRs, and reassign at least half of legacy address space.
You have until 2011 to get onto IPv6, or otherwise make it so you never
more IPv4. Aggressive outreach to encourage organizations to transition
IPv6 and/or make broad use of address translation. Barriers to IPv6
DATA USED FOR ASSUMPTIONS
There are 35 /8 unavailable, 91 legacy, 39 to be assigned.
Global demand is increasing since 2002, by very roughly 1.5 /8 per year.
It doesn't matter whether you measure IANA assigning the last /8 (or the
last 5 /8s) or RIR runout date. By 2011, RIRs will be
assigning/allocating 1.5 /8 per month, so the last ARIN assignment will
be only 3-6 months following IANA runout.
I'm a liberal arts guy, not a math guy.
This is all my own work, for my own use. I didn't check with anybody
I ignored data from 2008 (although run rate is about right).
I assumed that improved efficiency in legacy space would be the same for
Classes A, B, and C, though I believe this assumption to be flawed. Of
18 Class As are privately held, I believe 4 Class A holders could be
convinced to renumber, given enough money. Assuming a significant
Class B and C holders (let's say 30/44 Class B holders and 6/7 Class C
holders) can renumber, that's still only 40 out of 91 legacy /8s
Assuming they have only 25% efficiency, they use 10 /8s to renumber
giving us a net of 30 /8s
Not all RIRs assign address space at the same rate. If you want to
the burn rate on IPv4, APNIC and RIPE NCC are the areas of highest
consumption, followed by ARIN.
I've attached the spreadsheet I used, so you can check my work. I'll be
surprised if it survives scrutiny.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: IPv4 run rates.xls
Size: 39424 bytes
Desc: IPv4 run rates.xls
More information about the ARIN-PPML