[arin-ppml] Policy discussion - Method of calculating utilization

[Brett Frankenberger]

On Fri, May 02, 2014 at 09:05:06PM -0500, Jimmy Hess wrote:
> On Fri, May 2, 2014 at 8:52 PM, Brett Frankenberger
> <rbf+arin-ppml at panix.com> wrote:
> > Why is it not OK to get more space when you have an unused /21 that
> > is not adjacent to your other space, but it's OK to get more space if
> > you have an unused /21 hidden inside a /16?
> 
> > I support the proposal.
> 
> You assert both should be OK.   I assert neither should be OK,
> favoring the more rigorous justification criterion as better
> stewardship.

Actually, I assert that both should be the same.  I agree with Jeff's
point that, for example, 32 /21s are the same whether you got them at
one time as part of a /16 or on 8 separate occasions as part of 8
separate /19s ... and that we should fix that, and, separately, if
there are problems with the utilization requirement (for example, if
80% is not stringent enough) that should be handled separately.

> And whether each individual allocation has to be utilized or not, the
> calculation method, is inherently entangled with the utilization
> criterion.
> 
> It may be more work (more required renumbering or greater cost / more
> local router entries required)  to efficiently utilize the /21 hidden
> inside the /16,   in case this is not a contiguous  /21,   but a
> fragmented group of a few hundred /28s and /27s  spread around the
> entire  /16  due to  lots of number releases over time, or an
> ineffective allocation plan.

Well, sure.  And perhaps policy should take into consideration not only
the utilization percentage, but also the distribution of the unused
space.  For a number of reasons, I disagree with that, but my point in
this thread is that it's separate from question of whether or not we
should calculate utilization differently for two organizations that
have exactly the same amount of address space and exactly the same
utilization, when one org got its address space as a smaller number of
larger blocks, and the other got its space as a larger number of
smaller blocks.

If the right thing to do is to count smaller blocks of unused space
differently from larger blocks of unused space, that should be a
separate proposal.

(Note that the contiguous free /21 is not the common case here.  The
more common case that is relevant to the proposal at hand here would be
the most recently assigned space being 50% or so utilized -- and
perhaps as fragmented as you describe above -- while all space that the
organization has been allocated, in aggregate, is well over 80%
utilized.)

     -- Brett