On the RRG list:
http://www.ietf.org/mail-archive/web/rrg/current/msg04907.html
As part of an RRG discussion with Bill Herrin, I created the idea of
a "Distance Server" by which an ITR could choose (or have already
chosen by something else) the closest (in BGP terms) ETR of multiple
ETRs in the mapping.
Bill suggested that an ITR can already do this, but it can't.
LISP or Bill's TRRP could have this as a separate server, but it
could also be integrated into LISP-ALT's Map Resolver.
For Ivip, I would integrate it into the full database Query Server
QSD. For APT, the logical thing is to integrate it into the Default
Mapper. The same thing could be used for a core-edge elimination
scheme, but I think core-edge separation schemes are the best approach.
Without a "Distance Server" or similar, neither a core-edge
separation scheme nor a core-edge elimination scheme can perform the
equivalent of the existing, conventional BGP-based "anycast" in a way
which scales better than the existing BGP approach. This BGP
approach scales very poorly because it relies on a separate BGP
prefix for every such set of "anycast" servers. See:
http://www.ietf.org/mail-archive/web/rrg/current/msg04897.html
http://www.ietf.org/mail-archive/web/rrg/current/msg04904.html
With a "Distance Server", I think LISP, APT, Ivip and TRRP etc. could
do this work very nicely, in a highly scalable fashion.
This goes beyond conventional anycast and into high reliability and
"disaster recovery" systems, which Bill gave an example of and which
I quote in my RRG message.
- Robin
Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.