RIPE 43

Please mail comments/suggestions on:

• content to the Chair of the working group.
• format to webmaster@ripe.net.

Minutes Routing WG at RIPE43 (first draft)
===========================================
Tuesday September 10 2002, 14:00 - 17:00

Chair: Joao Damas
Scribe: Rene Wilhelm


A. Preliminaries (Joao Damas)
-----------------------------

Joao Damas welcomed everyone and explained that Joachim Schmitz,
chair of the routing working group, couldn't make it to the
RIPE meeting. Therefore Joao chairs this session.
Rene Wilhelm volunteered as the scribe.
Agenda and minutes from RIPE42 were approved.


B. Update from the RIPE NCC (Andrei Robachevsky)
-------------------------------------------------

The status of outstanding actions was discussed. Work at RIPE NCC is
ongoing. First phase (AS numbers) of early registration transfer done,
RIS' list of announcements in unallocated address ranges by RIPE44.

41.R1 on M.Blanchet, D.Kessens, F.Parent, J.Schmitz, J. Damas, RIPE NCC
Continue work on RPSLng
- ongoing - Presentation in second part of session

41.R2 on J.Schmitz, RIPE NCC
Continue work on RRCC
- ongoing - Standalone version now available.

42.R1 on RIPE NCC
Implement objects in database for unallocated IP address space
- open -

42.R2 on RIPE NCC
Arrange with other registries coverage of complete unallocated
IP address space
- open - depends on the above.

42.R3 on RIS team
Identify unallocated IP address space in RIS
- open -

42.R4 on RIPE NCC
IRRToolSet: pointers on webpage to implemented platform
- DONE -



C. Route Flap Damping: Harmful? (Randy Bush)
-------------------------------------------------

Slides are available (PDF format) from
http://www.ripe.net/ripe/meetings/archive/ripe-43/presentations/ripe43-routing-flap.pdf

Randy Bush presented results of a study on route flap damping,
inserting and withdrawing routes with a beacon and looking at
the effects at Oregon route-views and RIPE NCC RIS collectors.
Surprising result: due to cascaded withdrawals one single flap
(for example a reboot of a router) can cause a prefix to be
suppressed.

Route flap damping parameters may have to be revised, but
more data are needed; Randy asks for help, need more
BGP beacons installed.


D. RIS Update (Matthew Williams-Bywater)
-------------------------------------------------

Slides at:
http://www.ripe.net/ripe/meetings/archive/ripe-43/presentations/ripe43-routing-ris/

Matthew gave an update of the status and work done in the RIS project.

Statistics: in August the RIS saw a total of ~140000 different prefixes,
about 4800 hits on both web sites and nearly 750 MB of Raw data
downloads. Further analysis showed these downloads are primarily
done by university groups and research sites. People who use the
data and publish research papers are kindly requested to inform
the RIS team.

As of June 18th, the first route collector in North America is fully
operational. Located at MAE west, this rrc08 had 8 peers at the time
of the meeting; more are welcome (contact: rispeering@ripe.net)

New features on RIS website:

- AS and Prefix queries extended with Sort options
- New Plots added to RIS-report (AS types, AS path length)
- Daily generated lists of Martians and BGP Traffic Hot Spots added
- Daily plots from Agilent Labs' RIS Data Analysis Project,
visualizing and characterizing BGP behaviour.
- Help pages extended

Ongoing Projects:

- Improving the database insertion process, faster insertion,
store *all* BGP attributes
- Identifying Unallocated Address Space
- MyAS (concept): send notifications in case of unexpected routing behaviour

As an example of using the RIS, Matthew talked the audience through
the observations surrounding the shutdown of Ebone. Prefix queries
show the migration of Ebone networks to KPNQwest AS and the final
shutdown of these networks.



E. A RIS-based System for Visualization of Routing Flaps
(Maurizio Patrignani)
--------------------------------------------------------

http://www.ripe.net/ripe/meetings/archive/ripe-43/presentations/ripe43-routing-ris-flaps/

Maurizio presented the system developed at the third university of Rome
for visualizing routing instabilities. It runs as a Java applet
in a web browser. Main page: http://www.dia.uniroma3.it/~lombardo/flapviewer

Version 2.0 of flapviewer is expected to be released in March 2003,
it will feature a.o. improved visualization and efficiency.


Question: Is it possible to download and run the visualization program
on ones own computer?

Maurizio: Currently the tool must be used through the website, we did
not yet consider releasing the code for graphics to allow local
installations.


F. Observations on redundant BGP traffic and flaps from the RIPE RIS
collectors (Alexander Tudor)
---------------------------------------------------------------------

http://www.ripe.net/ripe/meetings/archive/ripe-43/presentations/ripe43-routing-redundant/

Alexander Tudor reported on the analysis of RIS data done at Agilent Labs.

The rationale of this work is to:

- Identify patterns in BGP traffic measurements
- Quantify performance cost of BGP processing
- Detect anomalies

Alex showed several graphs derived from January 2002 RIS data, both for
one full feed peer at LINX and one full feed peer at AMS-IX:
number of daily announcements, daily updates, duplicate announcements, flaps,
distribution of interarrival time of updates & announcements,
and flap duration.

conclusions from these data:
- 20% of all updates (~100k prefixes) concern duplicate announcements
- 35% of all updates have 30 second periodicity
- less than 3% duplicate withdrawals

Following this results from daily measurements were presented:
number of announcements, duplicate announcements, withdrawals, duplicate
withdrawals and number of flaps, all on hourly basis. These graphs
are generated daily and published on the RIS website.

Next steps in this analysis:
- attempt to identify spatial and temporal patterns; is there a baseline?
- develop methodology to quantify performance cost of BGP processing


Questions?

No questions from the audience, but Alex had some _for_ the
audience:

- when we compare peers what is right amount of time over which
they should be compared? 1 minute, day, 15min, hour?
something to think about.

- we want to make the effort to make this useful to the operators,
but are we taking the right measurements? what do _you_ want to see
measured on a daily basis?

If you think it is of use to you, let Alex know what you want.

Alex also called for volunteers who could help correlate
observed behaviour with known events, trouble tickets.


G. RPSLng, the new draft (J.Damas)
----------------------------------

http://www.ripe.net/ripe/meetings/archive/ripe-43/presentations/ripe43-routing-rpslng/


Joao presented the new draft for extending RPSL with capabilities to
describe multicast and IPv6 routing policies.

Questions:

Wilfried Woeber: is this an update to the RFC? what is the formal status,
the next step?

Joao: it is an update to the earlier draft, it fills in holes.
Want to massage it to internet-draft form and submit to wider community.

RIPE NCC will make it supported by the IRRtoolset.


Y. Input from other WGs

No input from other working groups.


Z. AOB

No other business.