Internet Standards/CD1.mdf/INET/IETF/BGPDEPL/93NOV.MIN

Text File | 1994-02-17 | 12.3 KB | 333 lines

CURRENT_MEETING_REPORT_


Reported by Peter Ford/LANL

Minutes of the Joint Session of the BGPDEPL Working Group and
CIDRD BOF

The BGP/CIDR deployment meeting was held on November 3, 1993 and was
chaired by Jessica Yu and Vince Fuller.  The first order of business was
a brief status report on BGP-4 implementations:


   o ANS (Guy Almes):  The ANS test network mid-November.  Deployment in
     the ANS production network during December.

   o cisco (Paul Traina):  In beta.  Get the image from ftp.cisco.com.
     Please join the beta list; mail to [e-mail].

   o Wellfleet (John Krawczyk):  Full product in 8.0 by Spring/Summer
     1994 test version in by January/February.

   o 3Com (Tracy Mallory):  Beta available.

   o BBN: Under development.

   o Europanet testing in progress, deployment by end of the year.

   o Rainbow Bridge (Rob Coltun):  Status?


Peter Lothberg and Andrew Partan reported on their BGP-4 test network.
It is a virtual test network which is accessible to anyone who wished to
participate.  The current players are predominantly cisco-based, and
many use cisco GRE tunnels to obtain connectivity with the test network.
There are currently 15K IP networks sloshing around.

The following participants are on the test network and ANS is expected
to be soon.


   o 3Com
   o Alternet
   o cisco
   o Ebone
   o ESnet
   o ICMnet
   o IIJ
   o NEARnet
   o RIPE NCC
   o STUPI


Peter Lothberg reports that he has converted the EBONE over to using
BGP-4 (nine routers).  Static aggregates have been injected into EBONE
from the regionals and passed over to other regionals and the ICM. He
also reported that the ICM system has also been cut over to BGP-4 (four
routers).  Andrew reported a similar cut over of the Alternet routing
system.

It was noted that the current BGP-4 code is beta code and one has to
carefully test their current configuration and operation prior to
deploying this code in full operation.  Join the beta list at cisco for
more details.

Andrew reported that Alternet uses IGRP within their system and Peter
Lothberg reports that he is using IS-IS.


Merit Routing Registry - Dale Johnson

From the network operators' point of view, there is a need to be able to
validate the aggregate routes received via CIDR. One approach is to
register aggregate routes in a database with its creator AS information
and the contact information of the AS could be obtained from various
existing databases such as the InterNIC, RIPE and Merit.  Merit, RIPE
and the InterNIC will work together on this.

Another approach mentioned at the meeting is to use BGP's AGGREGATOR
field to carry such information.


Guidelines for Block Assignment

Marten Terpstra gave a presentation on Guidelines for block assignment.
The RIPE people have worked with the European network community to build
a distributed operational model for Internet Registries (IRs).  They
currently work with a model of:  Global IR, Regional IR, Local IR. The
RIPE NCC allocates addresses to Local IRs based on the following
guidelines:


   o Get two year estimate of address usage.  Make sure blocks are
     CIDRable.  CIDR enforced to end sites.  Can claim unused reserve
     block.

   o The RIPE NCC has been assigning the CIDR way, since June 1992.  The
     procedures are documented in RIPE 72 which can be obtained via
     anonymous FTP or Gopher from ns.ripe.net.

   o The RIPE DNS scheme for 193.in-addr.arpa is to try to delegate
     zones to providers.  RIPE 84 documents policy (customer shifts,
     preserve mapping, etc.).

   o Dennis Ferguson asked about the current efficiency of use of IP
     address space from the European side.  It is estimated to be 3.8%.

   o RIPE only allocates for Europe and would like to advocate that
     regional registries do the same.

   o Marten noted that assignments do not really count as allocated
     unless they submit detailed information to the network, ensuring
     that the RIPE NCC is kept up-to-date.


BARRNet Allocation of Addresses

Jessica asked Vince to report on how Barrnet allocates addresses.  Vince
indicated that this is a manpower intensive process.  They sit down with
customers to get an estimate for two years out, then they work on a
subnetting scheme and do a crystal ball gaze.


InterNIC Allocation of Addresses

Mark Kosters of the InterNIC reviewed how they allocate addresses, which
was similar in spirit to the Barrnet and the RIPE NCC. Tony Bates, Mark
Kosters and Vince volunteered to write an allocation policy document
which can be used as guidance for providers.

The InterNIC will be doing block in-addrs.

When the InterNIC allocates a CIDR block to a provider, assignments are
requested to be sent back to the InterNIC who will register the
assignments to the InterNIC database.  The Shared WHOIS Project (SWIP)
is working on getting many of the InterNIC-related electronic updates
done in a timely manner.  The RIPE NCC does all their updates
electronically and are happy with the results to date.

There were several suggestions from the working group to the InterNIC
with regard to policy.

It is felt that the InterNIC needs to do more preallocation of blocks
for IRs that are to be delegated.

Marten suggested the InterNIC look into swapping CIDRable Cs for Bs that
are already allocated, but not heavily utilized.


``Greening of the Internet''


Vince led a discussion on ``Greening of the Internet.''

The outline of the talk was:


   o Class A usage with CIDR
   o Subnets and CIDR
   o Renumbering issues and tools
   o Politics of CIDR---block sizes, provider responsibility
   o CIDR Analysis discussion and question and answer


Class A Usage with CIDR has a few small problems:


   o There is a DNS issue, which relates to how providers delegate the
     in-addr namespace.

   o Dumb multihomed host problem.  Using older BSD systems as routers.
     It is felt that if a system can not handle variable length subnet
     masks they are obsolete.


VLSM needs to be better documented, and their use, made simpler by
better tools and education.  It is observed that most sites simply use
8-bit subnets since they are the easiest thing to read, use, etc.

Charley Kline who is in charge of networks at UIUC, described how they
allocate subnets.  Following methods described in RFC 1219, and
implementing software to help administer the address space, UIUC has
been able to manage 13K hosts and 324 subnets in a single Class B
network.  Charley illustrated the methodology using binary trees.

Tom Easterday and Charley Kline volunteered to work on a document
describing the use of VLSM and better utilization of subnets in a single
address block.  Havard Eidnes agreed to help and offered the use of his
INET '93 paper as a starting point.  Havard's paper is available in the
INET '93 proceedings via anonymous FTP and Gopher from
cnri.reston.va.us.

There was a unanimous vote in favor of using the IP addr/len syntax for
representing prefixes.

Scott Bradner stated that it is important for the ALE and CIDRD Working
Groups to establish goals and objectives for address space usage.

Vince presented a ``pain and anguish'' slide which went into the issue
of renumbering sites.  The discussion focused on better uses of the
already allocated Class A network addresses.  Vince pointed out that
CIDR does not require one to renumber when a site leaves one provider
for another, but to maintain a minimal state of routing information is a
good idea.  To support this activity we need tools and documentation on
renumbering.  The use of DHCP can help reduce the effort in converting
from one IP address block to another.

Barrnet has helped to renumber several sites and has used the following
procedure:


   o Add new DNS NS addresses at the InterNIC.
   o Add new addresses on the primary server, wait for propagation.
   o Reconfigure the network.
   o Delete old addresses from the primary server.
   o Delete old NS addresses at the InterNIC.


Barrnet customers have not had problems with renumbering, provided the
customers are given a good set of instructions.  Transitions must be
gradual if they are to work.  Secondary addresses are needed to
facilitate transition, and most routers support this.

Yakov Rekhter volunteered to discuss dynamic updates of DNS with the
Domain Name Systems Working Group (DNS).

Several other ideas to conserve address space came up during discussion
including ARP being changed to be like ES-IS for IP, dynamic prefix
updating, etc.

Tony Li put up a prototype charter of the proposed ALE Working Group.
The primary purpose is to watch utilization numbers.

Andrew Partan used Alternet data to show how much CIDR can buy you.  The
number of networks from AS701 shrinks from 2100 to 650 today.

There was a brief report on how big an Internet can be routed today:


   o 28-29K routes in a 16 MB cisco.
   o 25K routes in the ANS routers.


Tony Bates volunteered to help continue the monitoring mode.  Tony
Bates, Marten, David Conrad and Vince will document how to better use
address space within sites.


Attendees

Susie Armstrong          [e-mail]
Jules Aronson            [e-mail]
William Barns            [e-mail]
Tony Bates               [e-mail]
Erik-Jan Bos             [e-mail]
Rebecca Bostwick         [e-mail]
Jim Bound                [e-mail]
Rich Bowen               [e-mail]
Scott Bradner            [e-mail]
Al Broscius              [e-mail]
Jeffrey Burgan           [e-mail]
Enke Chen                [e-mail]
Henry Clark              [e-mail]
Michael Collins          [e-mail]
Rob Coltun               [e-mail]
John Curran              [e-mail]
Michael Davis            [e-mail]
Taso Devetzis            [e-mail]
Christopher Dorsey       [e-mail]
Tom Easterday            [e-mail]
Havard Eidnes            [e-mail]
Robert Enger             [e-mail]
Stefan Fassbender        [e-mail]
Dennis Ferguson          [e-mail]
Robert Fink              [e-mail]
Dale Finkelson           [e-mail]
Peter Ford               [e-mail]
Catherine Foulston       [e-mail]
Vince Fuller             [e-mail]
Vincent Gebes            [e-mail]
Herluf Hansen            [e-mail]
Denise Heagerty          [e-mail]
Matt Hood                [e-mail]
David Jacobson           [e-mail]
Dale Johnson             [e-mail]
Matthew Jonson           [e-mail]
Jeanine Kamerdze         [e-mail]
Akira Kato               [e-mail]
Hiroshi Kawazoe          [e-mail]
Sean Kennedy             [e-mail]
Charley Kline            [e-mail]
Mark Kosters             [e-mail]
John Krawczyk            [e-mail]
Walter Lazear            [e-mail]
Tony Li                  [e-mail]
Robin Littlefield        [e-mail]
Kim Long                 [e-mail]
Peter Lothberg           [e-mail]
Bill Manning             [e-mail]
Glenn Mansfield          [e-mail]
Jun Matsukata            [e-mail]
Stephen Miller           [e-mail]
Pushpendra Mohta         [e-mail]
Dennis Morris            [e-mail]
Jun Murai                [e-mail]
Michael O'Dell           [e-mail]
Vijayaragavan Pandian    [e-mail]
Andrew Partan            [e-mail]
Brad Passwaters          [e-mail]
Michael Patton           [e-mail]
David Piscitello         [e-mail]
Kenneth Rehbehn          [e-mail]
Yakov Rekhter            [e-mail]
Isil Sebuktekin          [e-mail]
Paul Serice              [e-mail]
Erik Sherk               [e-mail]
Timon Sloane             [e-mail]
Frank Solensky           [e-mail]
Bernhard Stockman        [e-mail]
Larry Tepper             [e-mail]
Marten Terpstra          [e-mail]
Claudio Topolcic         [e-mail]
Jerry Toporek            [e-mail]
Paul Traina              [e-mail]
Keisuke Uehara           [e-mail]
William Warner           [e-mail]
Chris Wheeler            [e-mail]
Jane Wojcik              [e-mail]
David Woodgate           [e-mail]
Jessica Yu               [e-mail]