[arin-ppml] Utilization policy is not aggregate
Jimmy Hess
mysidia at gmail.com
Sat Nov 17 12:07:56 EST 2012
On 11/16/12, Larry Ash <lar at mwtcorp.net> wrote:
> Reaching 80% on a smaller allocation is a lot
> harder than a /18. Over time holes develop in the utilization.
> You will reuse them but at any given time it's difficult if not
> impossible.
> How about 80% overall with no single allocation under 70% (whatever).
OK.... that solves that particular problem.
Although, still... the 20% criterion is really there to give
applicants time to request new IPs before their existing resources are
exhausted; I would say that "allowing holes" would be an
unintended side effect of the policy.
How exactly do we come up with the 80% rule in the first place, and
how exactly do we prove that 80% is the right number, and not 90% or
75%?
Instead of looking at the size of the allocation, and the number of
IPs used, perhaps policy for ISPs should look at the size of the
allocation, and the _average size of a suballocation_ of each size
from /32 down to the length of the block assigned.
e.g.
Total average number allocated per 12 month period
/32 0
/31 0
/30 0
/29 10
/28 5
/27 3
/26 2
/25 4
/24 1
And state, that new resources may be assigned when any of these
conditions are true:
(i) Of all IPv4 resources previously received, 99% have been suballocated.
(ii) The number of unallocated IP addresses sums to a number less than
or equal to the number of IP addresses allocated over the
previous 6 month period.
(iii) The ISP organization can show, that when the ISP makes
allocations for the next 3
months, it is not possible to map their existing unallocated
address space, onto
each of those allocations, without renumbering customers,
Or reducing the available aggregate IP addresses that can be
allocated below
the size of that ISP's average allocation, times the average 3
months' number of allocations of that size.
> The 80% policy has always greatly favored the big guys. It's much
> easier to reach 80% on a /18 than a /20. Natural holes that occur
> always seem to amount to almost 10-25% in a /20 unless much of it
> has been delegated as /23's or /24's.
--
-JH
More information about the ARIN-PPML
mailing list