[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RRG] Consequences of no renumbering...



Hi Lixia,

You wrote, in part:

> - there are multiple proposals on the RRG table that fall into the
> general map-and-encap direction. In general they do not require
> renumbering.

I think it would be best to clarify what is meant by "renumbering" -
at adoption of the new system or as a regular event during system
operation?

Once a network has some micronet space (Ivip) AKA EID space (LISP,
APT and I think TRRP) there is no renumbering of that space as
different ISPs are used, including when traffic is switched from one
ISP's ETR to another ISP's ETR.  Both involve a mapping change in
Ivip, but that is not in any sense a renumbering of the site's network.

For Ivip or any of the other potentially practical schemes - LISP,
APT and TRRP - to scalably provide full multihoming and portability
support for packets from networks without ITRs, the micronets / EID
prefixes of multiple (ideally thousands of) end-user networks need
to be contained in a single BGP advertised prefix.  In Ivip, this is
called a Mapped Address Block (MAB).  The same requirement exists
for LISP Proxy Tunnel Routers, but there isn't a name for this BGP
advertised prefix which covers many EIDs.

Renumbering to adopt the new system is a once-only change which
could be practical for many end-user networks.  Those with PA space
need to renumber anyway, just to get another ISP or to get
conventional PI space.

I wrote about the prospects of smaller PI networks and all PA
networks renumbering once to adopt the new "Scalable PI" (SPI) space:

  http://psg.com/lists/rrg/2008/msg02355.html

This is for Ivip, but I think the same principles apply to LISP with
PTRs or to however APT and TRRP provide support for packets from
non-upgraded networks.  That support can only be provided by devices
which advertise a prefix in the DFZ.  In order to scale well, they
need to advertise a much smaller number of prefixes than the number
of micronets they are serving.

If you have a thousand existing networks currently in disparate
parts of the address space - PI or PA - and they need to share one
or a few BGP advertised prefixes for LISP PTRs, Ivip OITRDs or
whatever, then it follows that all these networks need to renumber once.

So the statement above is only true of renumbering during the
operation of the new system.  Most networks will need to renumber
once to adopt the new system.


Also, while Ivip has map-encap modes for both IPv4 and IPv6, it is
no longer a purely map-encap proposal.

There are two new Forwarding modes of transporting packets from ITRs
to ETRs without encapsulation:

   ETR Address Forwarding (EAF) - for IPv4
   http://tools.ietf.org/html/draft-whittle-ivip4-etr-addr-forw-01

   Prefix Label Forwarding (PLF) - for IPv6
   http://www.firstpr.com.au/ip/ivip/ivip6/

Ideally, we could upgrade the DFZ routers to support these, so we
wouldn't need the greater complexities and lower efficiencies of
map-encap.  Alternatively Ivip would be deployed initially as
map-encap and transition to Forwarding only once the DFZ routers
were upgraded.

The Forwarding modes don't alter the renumber-once at adoption
requirement.

  - 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