FWIW:
The sub-optimal routing will only be for the LISP map-request.
The reply and all encapsulated data traffic is between RLOCs
The ALT is merely there to provide the correct mapping.

Also, since a map-reply from any ETR for a given site will
provide you with all the RLOCs for that site, there is no
need to advertise any sub-prefix into the ALT.  Rather you
would advertise the most highly aggregated route you have
for your EID-site.  So the ALT table should be considerably smaller.


On Jan 25, 2009, at 11:49 AM, Benson Schliesser wrote:

Dave-

On Sun, Jan 25, 2009 at 3:26 PM, David Meyer <[email protected]> wrote:
       This is just an artifact of the way I built it. You could
       put up a ALT tunnel betwen Savvis and L3 if that was of
       interest (and Savvis & L3 wouldn't have to further
       advertise those routes to the rest of the ALT). Also,
...
       Or am I answering the wrong question?

I think you're answering the right question from the LISP-ALT
perspective. It theoretically exists for other solutions in the same
class, too, i.e. solutions that create a stretched path in order to
follow address-assignment topology instead of underlying forwarding
topology.

In the LISP-ALT case, however, I'd like to explore your answer a bit
further. Does this suggest that the answer is for my ALT AS to peer
with all of my existing Internet peers, if I want to avoid suboptimal
forwarding? If I did so, how would my ALT routing table size compare
to my current routing table size? I'm imagining them the same size
(i.e. comprised of supernets learned by my peers as well as
less-specifics learned when my peers' customers multihome) but I might
be missing something.

Cheers,
-Benson
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to