CURRENT_MEETING_REPORT_
Reported by Gene Hastings/PSC
Minutes of the Joint Network Joint Management Working Group (NJM)
Below are the combined Minutes for NJM and Network Status Report.
Announcements: Due to the number of presenters, there was an extension
of Network Status Reports into part of the NJM session. Supplementary
notes for the NETSTAT portion, as well as presentation slides can be
found in Section 3 of the Proceedings.
There was positive feedback after both sessions for having combined
presentations and discussions spurred by them. Consequently the
sessions will be more tightly coupled in the future.
Mark Knopper, Jordan Becker - Merit/ANS T3 Network Status
A number of refinements and enhancements are planned, including RS960
FDDI deployment, and routing software changes such that an ENSS will not
announce 140.222 if it is isolated from the backbone.
Changes in backbone internal configuration are planned to minimize
coast-to- coast delay. Jordan said he will put the PostScript version
of the delay map on line. (also included in Proceedings)
Change tonight (17 November): CLNP will become encapsulated, [in one of
the T1 PSPs] and transferred over IP, instead of being switched in a
native stack. This is part of the migration plan to move all remaining
traffic from the T1 net to the T3. It will initially remain on T1 net.
General notes and announcements:
o EASInet at CERN will have an ENSS, connected to NY.
o Traffic Source/Destination pair statistics are sampled, with a
frequency of 1 in 50.
o Am map showing the T3 backbone with the T1 backup net is available
online. Question: Do you have priority queuing for management
traffic? - Not yet; AIX 3.2 will have priority queuing for routing
traffic.
o The MTU in the backbone is set to 1500 on most interfaces. With
deployment of new interfaces, many are being changed to 4000. ENSS
FDDI can do MTU discovery. The deployment strategy is designed to
avoid fragmentation on Ethernet interfaces, at the potential cost
of inefficient FDDI use in the short term.
o The dismantling of the T1 backbone circuits begins 12/2 Within CNSS
PoPs, there are DSU upgrades planned, granting multipoint
capability, and redundancy within PoP.
1
o Merit/ANS have issued an invitation to users, vendors and
developers to come to the ANS test net to test interoperability of
new BGP support.
Milo Medin - NASA Science Internet
An upgraded, year-round link to McMurdo base in Antarctica will be
through ARC2. It was a seasonal 56kb, and has been upgraded to full
time 384kb via IntelSat IV Lessons learned from NSI operations: Don't
encapsulate (DECNOT) on 56kb lines.
NSI Currently does DECNOT on TCP. Doing it over UDP was a lose.
DEC NSP is ``broken'' the way TCP used to be w.r.t. retransmission.
With DECNOT, TCP does retransmission, and NSP does not have to.
NSI network management is migrating to MSU (but not MCC) (``Polycenter
2000'')
Future efforts:
o Awaiting Commerce Dept. OK for ``Export of Internet'' to Russia to
allow network links to same and other parts of FSU.
o NASA participation with DoE ATM will have connections at Langley,
Lewis, Goddard, Ames, JPL - AGS+.
Bernhard Stockman, EBONE
There is now a security access scheme in EBONE routers (Amalgam of
Kerberos and TACACS). There will probably be a 256kb link from Bonn to
Stockholm.
Network Joint Management - 19 November 1992
Bob Collet - Sprint
Sprint operates three internets, with Internet connectivity.
1. SprintLink (Domestic US)
2. ICMnet-2 (Atlantic)
3. ICMnet-3 (Pacific)
Sprint uses cisco except in some private nets it uses Wellfleet. There
are nine domestic customers, and nine more being provisioned. Will be
demonstrating T3 P-P at COMNET '93 in February 1993. Most of the
routers are owned by Sprint. There is some customer owned equipment,
but only at customer premises. Sprint has management responsibility and
sole configuration control of customer router.
2
Rich Fisher - GSFC
Project group at Goddard Space Flight Center does data acquisition from
satellites, and redistributes it to numerous terrestrial labs. RAC :==
Remote Analysis Computer (All VAX) All links provided by PSCN (NASA
``phone co.'')
Tony Hain - ESnet
[There are no accompanying slides for this presentation.] The
Department of Engergy's (DoE) original high speed RFP was two years ago,
and began review in February 1992. An award was made to Sprint, with
TRW and cisco as subcontractor. A protest was filed so the procurement
is on hold. The General Accounting Office's (GAO) 90 day timer expires
December 23rd. The GAO will say nothing before then...Therefore there
will be a three month delay in deployment.
Initially access will be provided at T3, eventually a mix, up to OC-12.
All local loops will be fiber. It is NOT a managed router service. The
DoE and NASA will accept raw ATM. The router will be a cisco AGS+ with a
CSC-4 processor and HSSI interface for T3. The router will connect via
Digital Link CSUs. A new router will be needed to go beyond T3 rates.
The net will do IP and CLNP. Planners are trying to figure out DECNET
Phase IV, and negotiating with DEC to make Phase V genuinely be CLNP (as
opposed to being only close). They are still discussing which IGP to
use. (contemplating OSPF and IBGP).
Six DoE sites are planned as part of the initial project:
1. Lawrence Livermore National Laboratory
2. Lawrence Berkeley Laboratory
3. Los Alamos National Laboratory
4. FermiLab
5. Oak Ridge
6. Superconducting Supercollider, Waxahatchie, TX.
Continued plans call for the eventual connection of all ESnet sites, but
sites that are not currently connected to ESnet will not necessarily be
connected directly. If they were already being considered for a T1,
they may get a connection.
In addition, there are 5 NASA planned as well:
1. Langley Air Force Base
2. NASA Lewis Research Center
3. Goddard Space Flight Center
4. Ames Research Center
5. Jet Propulsion Lab
Mark Knopper asks: What is the plan and timeline for testing vs.
3
production?
o Service will be brought up off-line
o Several sites will be cut into ESnet
o Procurement includes an off-line testbed
Since NASA and DoE will be sharing a fabric, they will eventually move
their peering to this net (vs. at FIXes) There will be several logical
subnets on the fabric:
o DoE internal
o NASA internal
o DoE-NASA ``Phantom DMZ'' for peering.
SDSC has gotten funding from NSF to participate. Will initially be part
of the DoE subnet. Tony will have more technical information in January
(encapsulation, etc.).
Mark Knopper: ERNET.(India Research And Education Net).
ERNET folks visited several US operators and Internet researchers,
including Merit. Funded by Indian Government and the United Nations.
Phase III (refer to slides) uses VSAT for domestic networking due to
inadequacy of domestic telecomm IP and CLNP over X.25. 128kb up, and
512kb down
Original connection was Alternet UUCP, dialup uucp, analog leased uucp,
analog leased SLIP (and improved UUCP performance, as TCP is better at
line utilization) usually had > 20MB of mail in their queue.
Use mostly Cisco MGS, some CGS. Currently has 1 class B network,
nationwide...
NJM Discussion
o Route aggregation, BGP4 deployment can you do it?
o Does there exist a CIDR traceroute? does anyone know how to
interpret one?
o Dennis Ferguson says that as a transition plan, it is easier for
the midlevel to do route aggregation than the backbone. He
suggests that such clients advertise both explicit routes and an
aggregated route. The Backbone will install the explicit routes,
and announce only the aggregated one.
o Continued from BGP Deployment Working Group. (See also BGPDEPL
Minutes): Ref. Claudio Topolcic's timeline. Must examine
transition to CIDR - Tools? If operators do not do BGP4 by (date),
4
they will need to accept default only from backbone. What is that
date? If Merit's deployment plan works, it may be December 31,
1993.
o ROAD transition issues. Who really needs to do BGP4? Many
``stub'' or tail nets will not need to soon. What is the
operational impact of dinosaurs? Will they really die out? Can we
afford to sustain them?
Attendees
Henry Clark henryc@oar.net
Robert Collet rcollet@icm1.icp.net
John Curran jcurran@bbn.com
Tom Easterday tom@cic.net
Dennis Ferguson dennis@ans.net
Richard Fisher rfisher@cdhf1.gsfc.nasa.gov
Vince Fuller vaf@stanford.edu
Tony Hain alh@es.net
Eugene Hastings hastings@psc.edu
Mark Knopper mak@merit.edu
Kim Long klong@sura.net
Matt Mathis mathis@a.psc.edu
David O'Leary doleary@cisco.com
Andrew Partan asp@uunet.uu.net
Marsha Perrott mlp+@andrew.cmu.edu
Tom Sandoski tom@concert.net
Bernhard Stockman boss@ebone.net
Claudio Topolcic topolcic@cnri.reston.va.us
William Warner warner@ohio.gov
Evan Wetstone evan@rice.edu
Chris Wheeler cwheeler@cac.washington.edu
Linda Winkler lwinkler@anl.gov
Paul Zawada Zawada@ncsa.uiuc.edu
5