-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

>    * Are we concerned about routing table size?  Is there a cache
>      limitation? or
>    * are we concerned about routing entropy on the wire?  and/or
>    * are we concerned about other control plane entropy internal to the
>      router associated with routing? or
>    * are we concerned about something else?

I suspect it is the last--that this problem really revolves around
traffic engineering based on injecting longer prefix matches, rather
than the routing table size, strictly speaking. The problem is, at the
core, I think, that competing organizations have competing traffic
engineering requirements, and the primary way to express those competing
traffic engineering requirements is through the injection of longer
prefixes.

I think I've heard it said someplace that only about 1/4 of the typical
ISP's table is Internet routes, while the remainder are internally
injected or 2547bis, etc. Perhaps someone can confirm or deny this?
Based on Vince's presentation, about 50% of the "internet" routing table
is actually also injected for traffic engineer. Hence, we get to the
point where an overwhelming proportion of any given routing table is
there to engineer traffic flow, rather than to provide reachability.

If this is true, this puts a new spin on the set of possible solutions.

1. Provide some different form of TE that uses something other than
longest match routing.

2. Figure out where the routes injected for TE lose their usefulness,
and remove them at that point.

3. Other (more creative) ideas....

I'm not certain how a loc/id split would "solve" this issue, if this is
the issue.... One of the reasons I'd like to see a better problem
statement is to better understand the actual problem.

o Is it TE?
o Is it table size?
o Is the table size being driven by TE?
o Is the table size being driven by sheer numbers of hosts?

And then, beyond understanding that, I think it's important to look at
the tradeoffs when considering a loc/id split. I'm not saying it's a bad
idea, we just need to compare the doo we're stepping in to the doo we're
actually in before taking that step.

:-)

Russ

- --
[EMAIL PROTECTED] CCIE <>< Grace Alone

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGAnKdER27sUhU9OQRAg98AKC3AEd0ZxnI9I9BkhvCj1K5iNxVJwCeMEmm
qOpUooUP8d28Z/qZB+mI89A=
=0Dyw
-----END PGP SIGNATURE-----

_______________________________________________
Int-area mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/int-area

Reply via email to