Re: [RRG] Re: Fast and sparse mapping?

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

Short version:     Economic incentives for joining (aggregating)
                   multiple adjacent MABs (Mapped Address Blocks)
                   into one MAB.

                   In the future there may be annual fees for
                   advertising a prefix in BGP - any prefix, or
                   maybe just new prefixes.

                   Why a core-edge separation scheme such as Ivip,
                   LISP, APT, TRRP or Six/One Router can still
                   provide scalability benefits even if these
                   MABs (or whatever their equivalents are in
                   schemes other than Ivip) are not joined together.

Hi Brian,

You wrote:

> Robin argues:
> 
>> One way of improving scaling is for any of these million SPI 
>> networks which have adjoining address space to advertise their
>> two or more UABs as a single MAB, instead of each UAB being a
>> separate MAB.
> 
> True. But that is, to use the technical term, a crap-shoot, just
> as BGP aggregation of adjoining PI prefixes is a crap-shoot. It
> will be the exception rather than the rule; that's the nature of
> a swamp. There are no natural economic forces that provide
> incentives for aggregation.

There might be some economic forces making the joining of two
adjacent MABs into one - since there will probably be economies of
scale regarding provision of mapping updates and OITRD services.

At some point in the future, it is possible that there will be a
more restrictive arrangement for BGP advertised prefixes.  For
instance, they may cost a certain fee per year, by some widespread
or universal agreement by a number of large providers that they
won't recognise any new prefixes which are not in this paid-for
scheme.  The revenues could be distributed in some way to help pay
for the core routing system in some way.  Even if the money was
given to charity, it would have the important benefit of reducing
growth in the number of DFZ routes.

There are various technical and administrative problems to solve,
and perhaps such a system would be contrary to various restraint of
trade laws - but I think that it is a likely development some time
in the next 5 to 15 years.  When the tragedy of the commons looks
like getting bad enough, a scheme such as this will be seriously
contemplated and then implemented - as is the case with taxes on
activities which release greenhouse gasses.

If that was the case, and if if the scheme developed to the point of
making each existing DFZ route a costly annual expense, then there
would be economic incentives to join adjacent MABs into a single
larger MAB (shorter prefix MAB), since OITRDs (Open ITRs in the DFZ)
for that space could then be provided with a single DFZ route.
There would be disincentives too, since whoever runs the currently
multiple MABs would need to agree on how they would all work
together with a single RUAS (Root Update Authorisation Server)
company for mapping changes, and with this, or another, company for
the OITRDs service.

However, I agree that expecting there to be a lot of aggregation of
MABs (joining adjacent MABs into one) would resemble a crap-shoot.

I wasn't suggesting that there should be a million MABs.  That was
an example of things not working out well.  If there were at some
stage a million PI end-user networks all converting their space to a
MAB, then it would still probably be better than before, assuming
many such networks split their space into two or more BGP advertised
prefixes.

It would be much better if most of those million networks decided to
relinquish their current PI space and instead gain some (perhaps
smaller) SPI space from a MAB operating company.  The very biggest
end-user networks such as universities probably wouldn't do this,
but it would be best if the smaller ones generally got out of the
BGP business and used SPI space, via Ivip or whatever.

There would probably be significant economic incentive to do this,
since the network wouldn't need BGP expertise, or have to worry
about responsibilities which go with advertising a BGP prefix.
Also, there would be a clear incentive to adopt SPI space if there
was some fee-based scheme for any BGP advertised prefix.


> So I stick to my guns: we need to map the edge-swamp into a
> significantly smaller core-swamp, or nothing will change. 

I agree.


> Relying on aggregation at the edge hasn't worked for BGP, so why should
> it work for any form of EID/RLOC mapping (regardless of terminology)?

Even if every existing PI network converted their space to an MAB,
and if none of them joined them together (aggregation), then the
core-edge separation scheme, such as Ivip, LISP, APT, TRRP or
Six/One Router would still be a routing scaling solution, because:

1 - Without the scheme, many of the existing PI end-user networks
    would have been splitting their space and advertising it as
    multiple prefixes.  With the scheme, they only advertise one
    prefix in BGP - their MAB.

2 - Most or all new end-user networks which need or want portable,
    multihomable, address space will obtain SPI space from an
    existing MAB, by renting it from a MAB operating company.

    Without the scheme, many of these would have got their own
    conventional PI space and advertised it in BGP as one or more
    prefixes.

  - 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