U.S. - France Link Upgrade

[Steve Goldstein--Ph +1-202-357-9717 <]

Folks,

A few days ago, I copied many (hopefully, most) of you with a message to
our ICM project asking for a task statement to upgrade our link to INRIA
from 128 to 512 kbps, concurrent with its being re-designated a Renater
link and its being re-homed from Sophia-Antipolis to Paris.  I used that
form of addressing to get the word out in a hurry, as I was pre-occupied
with official business away from the office, and I did not want to delay
the task at the ICM (Sprint).  (Renater would like to have the line
fully installed by 15 October and operational by 24 October.) Now, I have 
a bit of "breathing space", and I can discuss the matter in a more relaxed 
manner with the CCIRN and other interested parties.

For many months, usage data have shown the NSFNET-INRIA link to be
running at full capacity (over 90% utilization) at peak usage hours.
Normally, that in itself would have spurred a move toward a bandwidth
upgrade.  However, NSF's [informal] policy is consistent with the NACCIRN
and Euro-CCIRN [informal] policy/agreements/goals of consolidating 
intercontinental links, especially in the North America/Europe area.
NSF became a supporting member of Ebone92, mainly because Ebone92 offered
an opportunity to consolidate links and to route to a European service
which would then take care of routing within Europe.  So, rather than
simply upgrading the capacity of the NSFNET-INRIA link, we sought a
solution that, in a broader context, would be supportive of the pan-
European routing opportunity offered by Ebone92.  In the same timeframe,
French networking was re-organizing within the Renater framework, and
Renater was talking with NSF about re-designating (with INRIA's consent)
the NSFNET-INRIA link as a Rentaer link to the U.S. via NSFNET.  Renater
had also become a supporting member of Ebone92, and has offered free 
access to the U.S. via an expanded link within the context of Ebone92.
Thus, by connecting to the Ebone Boundary System operated by Renater,
we can achieve objectives of providing adequate bandwidth for the
connection to France as well as doing it within the pan-European routing
context offered by Ebone.  Further, as seen from the U.S. side, we view
ther links to [current-or-future] EBS's in Stockholm, London and Paris
as an integrated set of links between U.S. and European infrastructure
that should include balanced capacity (the infrastructural portions of
each link will be 512 kbps, initially), and integrated routing and backup
(see also note on physical diversity, below).

So, I invite comment from CCIRN colleagues.  For your information,
Sprint has terminated its support contract for NOC services with Cornell
University, effective at the end of this month (or was it June?), and has
already re-homed the NORDUnet link to the Washington, D.C., area.  The same
will be done with the link to France.  Washington area operations will be
performed by Sprint's new SprintLink (its commercial Internet service) NOC,
which is staffed around the clock (7 x 24) (Sprint has acknowledged the
assistance of Cornell in helping it to implement the SprintLink NOC, by the
way; this has been done in the spirit of Internet cooperation.)  Given the
uncertainties about the continuation of service over the EASIgate link to the
U.S. (the T1 from CERN), attention will be given to maximizing physical
diversity of the fat pipes from Europe (NORDUnet and France now on PTAT-1;
London on TAT-9, and EASIgate on TAT-8).  Plans for alternate routing for
robustness in the U.S. are also under consideration, but may not be implemented
immediately.  The cost of the international portion of the link will be
shared between Renater and NSF for the immediate future.  Later on, as
the NSFNET reconfiguration and European operations (e.g., OU and post-
Ebone92 developments) and GIX implementations all progress, the funding
issues may be subject to re-evaluation.  But this would hold true for other
intercontinental links as well.

I expect that saying all this is much easier than doing it properly from
an engineering (and perhaps also from an institutional) point of view.
To that end, I have asked the ICM to coordinate the U.S. side of the
(upgrade + re-homing) with the appropriate *EPG groups as well as
the Ebone Action Team and Renater (and France Telecom) to provide
stable engineering solutions at all levels.  Happily, there is a 
large overlap of membership among these groups and a history of cooperation,
so engineering coordination may prove to be a smooth exercise [let us hope!].
I note that the ICM folks (Sprint) have already begun dialogs with various
European counterparts and that they will also begin coordination with the
FEPG and other concerned groups.  I shall assume that Ebone and Renater
folks will take care of European coordination activities.

Best wishes,

Steve G.