On Mon, Jul 25, 2011 at 4:40 PM, Noel Chiappa <[email protected]> wrote: > I'm not so sure - a small/medium size site might have two ITRs (or a small > number of them), and in those cases responsbility for covering the address > space could be divided between them, but many smaller sites will only have one > (their exit router).
I agree. However, if the solution for ITR cache churn is to "Hail Mary" the packets up to the native Internet (assuming native connectivity for that Address Family is even available) then these packets must find their way to an xTR that can handle the traffic, or they will be dropped. This must be the very powerful router (or cluster) that some have mentioned -- but you have to be able to build and sell it at practical size, power/heat properties, and price. So saying this traffic can simply be forwarded natively and handled by a PITR is not a solution unless there are practical PITRs with enough capacity. The fundamental problem of data-plane/FIB scaling remains the same. In addition, if there were no IPv6 connectivity available to the ITR, it could not do native forwarding with hopes of reaching a PITR. It's a good thing IPv6 will probably be deployed everywhere long before LISP is a practical concern, but this exemplifies the fact that the current design of LISP may not be able to live up to its promise of an all-purpose encapsulation scheme able to carry any protocol within the outer-headers, without native support for that protocol/AF to the site. -- Jeff S Wheeler <[email protected]> Sr Network Operator / Innovative Network Concepts _______________________________________________ lisp mailing list [email protected] https://www.ietf.org/mailman/listinfo/lisp
