Final transition steps


Sean Doran writes:

Um, are these April 21 disconnections going to be permanent,
or is this just a test of limited duration (six or seven
hours, say)?

If no one notices and there are no problems, they will be

If the turn-off is going to last a long time then the NSFNET
International Connectivity Project will need MERIT to
process and have installed, no later than this Wednesday, a
NACR with changes to about 16000 prefixes. We can prepare
the NACR this weekend.

If this is temporary does that mean we have to do 16000 NACRs
next week instead?

We will also need to know whether this is a full turn-off of
the NSFNET backbone service next Friday as it will affect
our immediate plans for turning up an ICM DS3 connection to
MAE-EAST+, effectively moving the timing of that forward a
few days.

This will be a semi-full turn-off of the NSFNET backbone service.
The interconnects will remain in place. Those interconnects
and the associated ASs are listed in a separate message to
the wider world.

(To get the ICM DS3 up and running on time for the 21st
(instead of before the 30th) we have some timing issues that
are pretty critical, namely provisioning the DS3 and
installing all the equipment. It is unclear that this is
possible on the MFS side (never mind the paperwork process,
which we probably will just have to ignore in the short run).
What is clear though is that current connectivity to
MAE-EAST is insufficient for making 21 April a happy day --
ICM is transiting SprintLink's (too full) DS3 to MAE-EAST,
and using the FIX to reach the fednets and the networks
which have not yet transitioned; the current DS3 can't
support moving 10Mbps from the FIX, and don't even mention
the ethernet that ICM still has today).

If there are serious problems, please contact merit-ie and the
on-call engineer will work with you (we don't do paperwork though).

I am not opposed to the idea, nor to the implementation of
this push towards finding alternate connectivity. However,
I'm pretty annoyed that we got very short notice of what
appears to be an early shutdown of the NSFNET backbone
service. This is especially irritating as a shutdown on the
21st will dramatically affect routing worldwide, and force a
rewrite of a number of transition schedules which have been
in place for quite some time.

However, let's run with it. We'll know early in the week
when Peter, Vadim and I have spent alot more time on
moving ICM transition forward how bad things will be
at 9am on the 21st as opposed to the evening of the 26th,
and then again late at night on the 30th.

Right now, I humbly request that MERIT wearing its RA hat
quickly (i.e., this weekend, and certainly no later than
Wednesday morning) locates an RS at FIX-EAST so that ICM can
more rapidly bring up safe peering with the FIX-EAST
connectees than we had initially planned (i.e., we need to
have functional peering and transit set up this week instead
of using this week to debug things with the fednets who are
facing disconnectivity as of the 30th).

We love to provide RA service everywhere and anywhere - it's just
a matter of somebody contracting for the service. Any volunteers?

This RS should be seeded from the ENSS145 configurations
rather than any other source.

Having something in place at FIX-EAST which does the same
filtering that the AS 690 PRDB mechanism did has now been
made much more urgent than it was yesterday, so that we can
avoid rushing direct peerings, which traditionally has
resulted in serious routing botches that the whole world

Thanks for the vote of confidence.