Internet Info 1997 December/Internet_Info_CD-ROM_Walnut_Creek_December_1997.iso/ietf/rolc/rolc-minutes-94mar.txt

Text File | 1994-06-04 | 9KB | 245 lines

CURRENT_MEETING_REPORT_

Reported by Joel Halpern/Network Systems Corporation

Minutes of the Routing Over Large Clouds Working Group (ROLC)


Agenda

   o Introduction
   o Host resolution vs.  router resolution
   o NBMA ARP and NBMA NHRP
   o Dave Katz's changes to NBMA NHRP
   o Introduction to how we can use BGP
   o Joint meeting with BGP Working Group


First Session

Joel Halpern presented the working group's history and summarized the
problem statement and where the group currently stands.

Juha Heinanen discussed his Non-Broadcast Multiple Access (NBMA) ARP
draft document.

Dave Katz led a discussion of his update to the NBMA Next Hop Resolution
Protocol (NHRP), and the working group compared it to the NBMA ARP
document.  The working group discussed a number of possible technical
problems or pathological cases that might occur with the drafts,
including suboptimal routing and/or routing loops.

Joel led a discussion of whether the working group can restrict NBMA ARP
servers to be routers.  This was agreed following the discussion.

There was a protracted discussion, without resolution, of whether the IP
``mask and match'' restriction should be relaxed (and, on a very related
issue, whether hosts should ``ARP for everything'' [proxy ARP]).
Opinions on all sides of the issues were stated.

There was a non-productive discussion of which layer should be used for
the query protocol, IP or ICMP.

After some additional detailed technical discussions, the session was
adjourned.


Second Session

It was clear from the previous session that there was not a common set
of requirements and goals upon which technical solutions were being
based.  The working group appointed two editors of a Requirements and
Goals document (Caralyn Brown and Dave Katz), and produced a starting
set for the document.

Proposed topology:  very large NBMA networks with hosts, small routers,
large routers, and route servers attached.  The following cases are
explicitly included:


   o A small number of routers with many hosts behind each
   o Many directly connected hosts
   o Many routers with small subnets behind each
   o Many routers with many hosts behind each


Requirements:


   o Support for complex cases should not unduly burden simple cases.

   o Router-to-router forwarding must be loop free.

   o Must apply to multiple media and routing technologies of interest.

   o No new infrastructure deployed in order to support this.

   o This is applicable to other internetworking protocols (and not just
     IPng).

   o Support policy restrictions.

   o Support routing hierarchies and aggregation.

   o Be compatible with existing routing protocols.

   o Enable communications between two entities in different IP subnets
     on the same cloud.

   o Operate in the presence of multiple layers of routing aggregation,
     including aggregates which cross the cloud boundary.


Goals:


   o Do not hold up the simple case to solve the complex case (time to
     market).

   o The address resolution should be effective and cheap in terms of
     network resources.

   o Be decoupled from specific infrastructure, both media and routing,
     as much as possible.

   o The solution should converge to desirable routes.

   o Be extensible to handle QOS (and other new features) in the future.

   o Be able to determine the path before having sent any data packets
     over it.

   o Should be diagnosable.

   o Should be deterministic.

   o This should work on a cloud where not all attached routers are
     NHRP-capable.

   o Reasonable robustness in the face of misconfiguration.

   o Non-goal:  the working group is not trying to couple this work to
     link layer routing (this would involve significantly more
     complexity).


There was a discussion of whether to support network layer address
resolution to more than one link layer address.  There was no clear
agreement.

In preparation of the joint session with the BGP Working Group, Joel led
an overview of how the problem space is affected by using address
aggregation, and how BGP could be used to solve the resulting problems.

The working group met jointly with the BGP Working Group, and Curtis
Villamizar presented his BGP attribute draft document.  A number of
comments were received.  The BGP Working Group also informed the ROLC
Working Group that BGP4 is most probably frozen at this point, and
future enhancements may have to be a part of BGP5 (IDRP for IP).


Actions for Toronto

   o Caralyn Brown and Dave Katz will produce the requirements draft.

   o Dave Katz will also edit the next version of NHRP/NARP.

   o Curtis Villamizar will produce a new version of his draft.

   o The value of an additional framework document will be discussed on
     the mailing list.


Attendees

Edward Allen             [e-mail]
Anthony Alles            [e-mail]
Susie Armstrong          [e-mail]
William Barns            [e-mail]
Stephen Batsell          [e-mail]
Nutan Behki              [e-mail]
Ute Bormann              [e-mail]
Scott Brim               [e-mail]
Al Broscius              [e-mail]
Caralyn Brown            [e-mail]
Steve Buchko             [e-mail]
Joesph Burrescia         [e-mail]
Ross Callon              [e-mail]
Greg Celmainis           [e-mail]
Brett Chappell           [e-mail]
Luo-Jen Chiang           [e-mail]
Bob Cole                 [e-mail]
Michael Collins          [e-mail]
Charles Combs            [e-mail]
Terry Davis              [e-mail]
Farokh Deboo             [e-mail]
Dario Ercole             [e-mail]
Dennis Ferguson          [e-mail]
Shoji Fukutomi           [e-mail]
Eugene Geer              [e-mail]
Shawn Gillam             [e-mail]
William Gilliam          [e-mail]
Michael Goguen           [e-mail]
Dragan Grebovich         [e-mail]
Chris Gunner             [e-mail]
William Haggerty         [e-mail]
Stuart Hale              [e-mail]
Joel Halpern             [e-mail]
Dimitry Haskin           [e-mail]
Ken Hayward              [e-mail]
Ian Heavens              [e-mail]
Juha Heinanen            [e-mail]
Greg Hill                [e-mail]
Eric Hoffman             [e-mail]
David Jacobson           [e-mail]
Ronald Jacoby            [e-mail]
Dale Johnson             [e-mail]
Matthew Jonson           [e-mail]
Merike Kaeo              [e-mail]
David Kaufman            [e-mail]
Manu Kaycee              [e-mail]
Sean Kennedy             [e-mail]
Berry Kercheval          [e-mail]
Percy Khabardar          [e-mail]
Ted Kuo                  [e-mail]
Mark Laubach             [e-mail]
Fong-Ching Liaw          [e-mail]
Paul Lu                  [e-mail]
Jamshid Mahdavi          [e-mail]
Andrew Malis             [e-mail]
J. Scott Marcus          [e-mail]
Jun Matsukata            [e-mail]
Keith McCloghrie         [e-mail]
Keith Mitchell           [e-mail]
Donald Pace              [e-mail]
Maryann Perez            [e-mail]
Drew Perkins             [e-mail]
Venkat Prasad            [e-mail]
Rex Pugh                 [e-mail]
Thomas Pusateri          [e-mail]
Murali Rajagopal         [e-mail]
K. K. Ramakrishnan       [e-mail]
Kenneth Rehbehn          [e-mail]
Tony Richards            [e-mail]
Robert Roden             [e-mail]
Duncan Rogerson          [e-mail]
Dan Romascanu            [e-mail]
Michal Rozenthal         [e-mail]
Timothy Salo             [e-mail]
Hal Sandick              [e-mail]
John Scudder             [e-mail]
Joshua Seeger            [e-mail]
Uttam Shikarpur          [e-mail]
David Shur               [e-mail]
Frank Solensky           [e-mail]
Mark Swanson             [e-mail]
Dan Tappan               [e-mail]
Dono van Mierop          [e-mail]
Curtis Villamizar        [e-mail]
Maria Vistoli            [e-mail]
Justin Walker            [e-mail]
Dan Wood                 [e-mail]
Jessica Yu               [e-mail]