ip-precedence for management traffic

David Hiers hiersd at gmail.com
Wed Dec 30 13:54:41 UTC 2009


> Totally out of the box, but here goes: why don't we run the entire
> Internet management plane "out of band"

This has been one of my favorite conversation-stoppers for years.  The
PSTN fought tooth and nail against the need for OOB control, but
2600hz was a problem that they could not solve, so they bucked up and
paid for a control plane.  Where do you think we'd be now if
Phreakers, Inc. still had access to a PSTN with an audio frequency,
inband control plane?

Don't we insist on, and brag over, data/control seperation within our
devices?  Isn't it groovy when a frame is never seen by the switch's
CPU/SUP?

Sure, I'm streching the analogy a bit here to make a point: many of
our problems arise from giving bearer traffic access to the control
plane.

If the world wants an internet that is as predictable and reliable as
the PSTN, it'll bear the cost of protecting the control plane.  A
fundamental choice in the protection scheme is physical architecture.
IB or OOB, it's always a good thing to be explicit in design
decisions, and not adopt legacy/heritage decisions without
consideration.



David

PS: If you want OOB access to your gear when your core switches freak
out, don't let those switches touch any frame between your WAN link
and console ports.




More information about the NANOG mailing list