CURRENT_MEETING_REPORT_
Reported by Guy Almes/Rice
TEWG Minutes
The TEWG met for a single session on Wednesday morning, August 1st.
Scott Brim, Chair of TEWG, was unable to attend and asked Guy Almes to
chair the session in his place.
The session focused on sharing information about three increasingly
important areas of Internet topology: Europe, the Pacific, and the new
Army Supercomputer Network (ASnet).
Rudiger Volk, of the University of Dortmund and a participant in RIPE,
led a presentation and discussion of connectivity both between Europe
and North America and within Europe. Olivier Martin of CERN also
contributed to the discussion.
Rudiger first focused on trans-Atlantic connectivity. Among the most
important links are the following:
o The 64kb/s line from EUnet at CWI in Amsterdam to the UUnet site in
Virginia. This serves the EUnet community directly and serves as a
backup for other nets, e.g., NORDUnet.
o The 64kb/s line from NORDUnet in Stockholm to the NSFnet site at
JvNC. This serves the NORDUnet community directly.
o The T1 line from CERN in Geneva to the NSFnet site at Cornell
University. This serves the EASInet community directly.
o The 56kb/s line from INRIA near Nice to Princeton University. This
serves users within France directly.
o The 56kb/s satellite line from DFN/WIN in Garching to the ESnet
site at Fermilab. This serves the DFN/WIN community within Germany
directly.
o The 9.6kb/s line from Karlsruhe to NYSERnet serves another
community within Germany.
Further, there is a planned upgrade of the DFN-to-ESnet line to use one
of the two `fat pipes'.
Rudiger and Olivier mentioned two problems that lead to asymmetric and
sub-optimal routes to Europe:
o The heavy use of default routes within some parts of Europe often
result in asymmetric routes in which packets go from North America
to Europe via some explicit route, while return packets use a
default path. Increased deployment of dynamic routing within
Europe should improve this situation.
1
o The use of MX records for some European sites cause very suboptimal
routes to be taken in some cases.
RIPE is working with others to help solve these problems.
There was a brief discussion of the situation in Britain. As a general
rule, IP traffic from the outside world enters the UK via an
application-level gateway in London, and is transmitted via JAnet using
the Coloured Book protocols to individual campuses. There are several
exceptions to this that we discussed. First, the University of Kent at
Canterbury is on EUnet, and thus connects to Europe and thence to North
America via CWI in Amsterdam. Also Milo Medin reported that, as part of
the `fat pipe' to London, some British sites will be served by JAnet
using `IP-over-X.25' techniques; this should be an improvement over the
current use of the JAnet application-level gateway.
Rudiger closed with some thoughts on a possible outline for an
intra-European backbone. He noted that currently, the three most
important trans-Atlantic lines are those at:
o CWI in Amsterdam,
o The NORDUnet hub in Stockholm, and
o CERN in Geneva.
There are plans upderway to upgrade the bandwidth of lines from
Stockholm to Amsterdam and from Amsterdam to Geneva.
Milo Medin, of NASA, reported on a recent meeting of PACCOM, which
coordinates the Internet within the Pacific Rim. At the present, there
is a 512kb/s terrestrial line from NASA/Ames to Hawaii, which serves
Hawaii and the following other sites:
o Japan via four 64kb/s terrestrial circuits. There is some work to
combine these to a single 256kb/s circuit.
o Australia via a 56kb/s satellite circuit. There is some work on
increasing the bandwidth of this circuit within the year.
Unfortunately, it will be quite some time until the circuit can be
converted from satellite to terrestrial.
o New Zealand via a 14kb/s analog circuit. There is some work on
using better modems, and possibly real-time compression boxes, to
increase the effective bandwidth of this line.
o Korea via a 56kb/s circuit.
Among the coming developments are the following:
o The possibility of a 64kb/s line from Japan to Europe. This would
2
complicate routing within the Pacific. The effective use of the
current low-speed lines is eased by the ability to use default
routing heavily from Pacific Rim countries to Hawaii.
o Discussions of adding Singapore and Taiwan.
In response to a question about networking to sites in Antarctica, Milo
expressed regret over current technical problems that prevent the
placement of a geostationary satellite there. More seriously, he
mentioned that work is being done on networking to Antarctica.
Bob Reschly, of ASnet and BRL, reported on the ongoing deployment of
ASnet, which serves the Army supercomputer centers and other Army labs.
The initial topology is a mixed T1/56kb/s topology centered at BRL.
ASnet is 138.18.
Connectivity to NSFnet is primarily through the ASnet site at the
Minnesota Supercomputer Center via MRnet and CICnet. A secondary
connection from the ASnet site in Vicksburg to the SURAnet site at
Jackson, Mississippi and through SURAnet is planned.
Several ASnet sites are also on MILnet, and a subset of these will be
used to route traffic between ASnet and MILnet. An ASnet router at
FIX-Ease would improve connectivity both to MILnet, to NSFnet, and to
other parts of the Internet.
One interesting technical aspect of ASnet is its planned use of crypto
equipment on all serial lines.
ASnet is openly connected to the rest of the Internet, and is to be used
only for science/research uses within the Army.
Attendees
Guy Almes almes@rice.edu
William Anderson wda@mitre-bedford.org
Jeffrey Burgan jeff@nsipo.nasa.gov
Eric Carroll eric@utcs.utoronto.ca
Rob Coltun rcoltun@trantor.umd.edu
Dennis Ferguson dennis@gw.ccie.utoronto.ca
Dale Finkelson dmf@westie.unl.edu
Vince Fuller fuller@jessica.stanford.edu
Hellmut Golde golde@june.cs.washington.edu
Michael Grobe grobe@kuhub.cc.ukans.edu
Phill Gross pgross@nri.reston.va.us
Michael Hrybyk mwh@educom.edu
Steven Hubert hubert@cac.washington.edu
Dan Jordt danj@cac.washington.edu
Kathy Kerby kkerby@bbn.com
Dan Long long@bbn.com
Olivier Martin martin@cearn.cern.ch
3
Matt Mathis mathis@pele.psc.edu
Milo Medin medin@nsipo.nasa.gov
Paul Mockapetris pvm@isi.edu
Philippe Park ppark@bbn.com
Robert Reschly reschly@brl.mil
Ron Roberts roberts@jessica.stanford.edu
Ken Stetten kstetten@nrao.edu
Roxanne Streeter streeter@nsipo.arc.nasa.gov
Rudiger Volk rv@informatik.uni-dortmund.de
Tom VonDeak tvondeak@nasamail.nasa.gov
Carol Ward cward@spot.colorado.edu
John Wieronski john@osc.edu
Dan Wintringham danw@igloo.osc.edu
Robert Woodburn woody@saic.com
4