[RRG] Speculative knee in DFZ growth curve - subdivision of prefixes

[Robin Whittle <rw@firstpr.com.au>]

In "Re: FYI -- Informal LISP BOF scheduled for lunch time on
Thursday" Tony Li wrote:

> However, some speculate that we will see a knee in the number
> of prefixes that are carried in the DFZ.  In these scenarios,
> as a market for v4 prefixes is established, it will become
> easier for organizations to acquire PI prefixes and inject them.
> Further, existing space holders will tend to de-aggregate them
> so as to maximize their market positions.  This combination
> could possibly lead to a knee in the size of the overall table
> and a resulting knee in overall Internet convergence time.
>
> I should stress that this is all speculative at this point and
> we will simply have to see how it all plays out.

Whilst standing upon a railway track and observing, visually and
audibly, an approaching freight train, strictly speaking, the matter
of it colliding with oneself remains speculative until observations
confirm that this has in fact occured.

Likewise the robust theories about increased carbon dioxide and
methane leading to climate change remained "speculative" until the
observations of change and damage could no longer be reasonably
disputed.

IPv6 can't reduce the need for IPv4 space in the foreseeable future.
 People will break up existing IPv4 prefixes into smaller ones so
more folks can inhabit the crowded island.  Right now, we are still
clearing the last patches of forest, but when that is gone, the only
way of using the space more efficiently will be subdivision.

With BGP, this subdivision will clearly lead to much worse growth in
the number of advertised prefixes.  Whether and when that would lead
to a "collapse in routing" depends on how "collapse" is defined.  I
suspect the costs of running the BGP system scale fairly smoothly
with the number of prefixes, amongst other things - but there is
certainly going to be stronger growth in the demand for advertised
prefixes in a few years time, as long as BGP is the only way of
providing multihomable space for end-user networks.

If we can get a good ITR-ETR scheme in reasonably wide adoption by
2013 or so, then this can do the slicing and dicing.  Furthermore,
end-users would then have an alternative form of multihomable
address space and so will be less tempted to get traditional PI
space.  The barriers to gaining and perhaps keeping PI space could
then be reasonably raised in order to reduce the global burden of
advertised prefixes.

As long as we rely only on the current BGP system, these problems
remain unsolvable, since each constrains the others from being fixed:

  Growing demand for multihomed end-user networks.

  Increasingly unacceptable burden on all DFZ routers
  due to growth in number of advertised prefixes - cost of
  routers and general concern over complexity and stability.

  New and current end-users satisfying their future space
  requirements by gaining more numerous, non-contiguous,
  prefixes subdivided from older not-so-heavily used
  prefixes.

It seems the only incrementally deployable solution (no host
changes, changes to minimal number of routers) is an ITR-ETR scheme
along the lines of LISP, eFIT-APT, Ivip or TRRP.


  - Robin


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg