RE: [sig-policy]Discussion document: Application of the HD-Ratio to IPv4
> I have no objections to the idea, but the logic of large ISPs
> requiring lower utilization rate hasn't sucken in to me yet.
I will try to answer your queries here, but I also hope that the explanation
will become clearer with my presentation at the meeting next week.
> I assume network subdivion issue was addressed by considering
> allocated addresses to to be utilized for sub-allocations. In
> Japan, the size of allocations are not always proportionate
> to the number of assignments either.
Through discussions with large ISPs I have found that heirarchical
management of a large address space is almost always desirable and necessary
to some degree. Large ISPs may wish to subdivide address space according to
the major service divisions that they operate, because those divisions may
be autonomous and have completely separate structures; they may wish to
further subdivide according to major regions, countries, provinces or states
in which they operate; and they may wish to subdivide according to POPs,
customer groupings, or other factors. A single ISP may wish to utilise
different modes of subdivision depending on their internal architectures and
administrative structures, which are difficult to predict; however the
predictable fact is that with large address spaces, the need for
hierarchical subdivision increases. This is of course the basis of the
argument for the HD ratio.
I am not arguing that these hierarchies actually do exist, because the
degree to which an ISP can actually implement such hierarchies is highly
constrained by the 80% utilisation requirement. Where heirarchies are
created, management of those structures to the 80% utilisation level is
generally very difficult; and where hierarchies are not created, management
is also very difficult, but for other reasons.
>
> What is the biggest motivation for applying AD-ratio instead
> of lowering a 80% fixed rate for all LIRs?
As argued in RFC3194 and 1715, an HD-Ratio value represents a consistent
level of management complexity (or "pain") regardless of the size of the
address space. A fixed percentage value on the other hand can represent a
very low level of complexity (for instance achieving 80% of a /24 is
trivial) or it can represent a very high level of complexity (achieving 80%
of a /8 is arguably extremely difficult).
It seems to me that there is no necessity to lower the current 80%
utilisation requirement for a /20 allocation, while there is a strong case
for lowering the requirement for much larger allocations. Between these two
extremes, there is a "sliding scale" of need, and the AD Ratio offers a
means to calculate that scale on a fair and rational basis.
I hope that helkps.
Thanks.
Paul.