[LINK] SRAM-based IP forwarding (flat routing) for IPv4 & IPv6

[Robin Whittle]

There is a crisis in routing and addressing - in IPv4 and IPv6.  The
Internet Architecture Board ran a workshop on this in Amsterdam in
October 2006:

   http://www.iab.org/about/workshops/routingandaddressing/

Firstly there is the shortage of IPv4 addresses.  Secondly there is
the burden placed on transit and multihomed border routers by the
increasing size of the "global BGP routing table".

I wrote an Internet Draft (work in progress) to make the next
generation of routers in the Default Free Zone (transit routers and
multihomed border routers) use a 72 Mbit Static RAM in each
interface, to directly store the Forwarding Equivalent Class for all
14,6880,064 /24 prefixes.  A similar arrangement would work for a
restricted range of IPv6 global unicast addresses:

  http://www.firstpr.com.au/ip/sram-ip-forwarding/
  http://tools.ietf.org/html/draft-whittle-sram-ip-forwarding-01

This would make the DFZ a flat routing system, down to the
granularity of 256 addresses for IPv4, with each router instantly
forwarding packets without any fuss, on boundaries of /24, no matter
how disaggregated the address space is.

There is a mailing list carrying on the discussions of the Amsterdam
workshop:

  http://www1.ietf.org/mailman/listinfo/ram

The main topic of discussion has been a proposal called LISP
(Locator/ID Separation Protocol) which attempts to keep the global
BGP routing table relatively small, by having special headers and
LISP tunnel routers at various points around the edge of the
network, which tunnel packets between themselves and so keep most of
the IP addresses used by applications out of the BGP system.

  http://tools.ietf.org/html/draft-farinacci-lisp-00

My proposal is also being discussed.


I finalised my ping-based attempt at estimating the utilisation of
IP addresses:

  http://www.firstpr.com.au/ip/host-density-per-prefix/

   - Robin