RIPE 49

Archived This content has been archived and is no longer actively maintained.
RIPE Meeting: 49
Working Group: Test Traffic
Status: Final
Revision Number: 1

Minutes of Test Traffic Working Group at RIPE49
================================================

Location: Renaissance Hotel, Manchester, UK.
Chair(s): First Session - Henk Uijterwaal,
Second Session - Alex Tudor
Scribe: Matthew Williams
Jabber: Willem-Jan Goudsblom

The presentations are online at
http://www.ripe.net/ripe/meetings/ripe-49/presentations/#tt

The RIPE 49 Webcast Archives can be found at
http://www.ripe.net/ripe/meetings/ripe-49/sessions-archive.html


Session 1 ---------
Date: Thursday, 23rd of September, 2004
Time: 09.00 - 10.30

A. Administrative Matters

- 09:05:08 -

The session was formally opened, followed by a few welcoming words
from the chair. After a short introduction and some ceremonious
agenda-bashing, the minutes from the previous RIPE meeting were
approved and the participants' list was circulated among attendees
without further ado.

B. Presentations

- 09:08:33 -

1. "Modeling of end-to-end delays using Weibull-distribution mixtures"
Jose Hernandez & Iain Phillips, Computer Science, Loughborough
University, UK

Comments:

Q: Did you try to relate the different modes inside your mixtures of
Weibull distributions, because each mode might be related to the state
of the network?

A: No, not yet. That is the next step; establishing what the mix of
distributions means. We also need to find out what the optimal number
of distributions is.

Q: The Weibull distribution has many parameters, wouldn't another
model be better? A: It only has two parameters, and we think it is a
good model that indeed relates to performance somehow. These
parameters really do reflect performance of a link.

- 09:34:38 -

2. The "Metropolis" Project Kave Salamatian, University Pierre and
Marie Curie-LIP6, France

Comments:

Q: What did you miss on RIPE NCC test boxes?

A: No problems, they are very good measurement probes. However,
management doesn't scale well in large networks. The main advantage of
the TTM project was its flexibilty when 'embedding' it into another
architecture. Research and operations have different needs.

- 10:14:26 -

3. TTM Status Update Henk Uijterwaal, RIPE NCC

Comments:

Q: [Slides #15] At which forums were g/ccTLDs approached?

A: At a few different ones, e.g. the CENTR meeting.

Q: Metropolis have developed framework for monitoring, partly focusing
on a more accurate alarm system. We should probably share ideas.

A: Great idea. Please send me some pointers.

- 10:37:02 -


Session 2 ---------
Date: Thursday, 23rd of September, 2004
Time: 11:00 - 12.30

B. Presentations (continued)

- 11:02:51 -

4. ?SMARTxAC?: A passive monitoring and analysis system for high-speed
links Pere Barlet & Jordi Domingo-Pascual, Universitat Polit?cnica
de Catalunya, BARCELONA Catalunya (Spain)

Comments:

Q: Does your system automatically understand traffic; for instance,
interpreting dynamic ports during the initialisation of connections?

A: No, ports need to be configured manually.

- 11:32:40 -

5. "CoMo", The Open Monitoring Infrastructure Project Gianluca
Iannaccone, Intel Labs, Cambridge, UK

Comments:

Q: It is great that Intel are interested in building a monitoring
platform, saves reseachers from doing it.

A: Reseachers are more interested in analysis. We want to build a
modular system, where the modules can easily be written and integrated
into the monitoring platform.

Q: What is Intel's business model for this project?

A: Firstly, we need 'buy in' from the community. Peer review and
information exchange are essential when building an IDS or similar.
Secondly, there are many start-ups in the field and Intel needs to be
at the forefront of developments. The important question is the
abilities of a 'network processor'; what primitives should it look at,
what should be provided, and how can the design be optimised?

- 12:02:13 -

6. 6QM Solution for IPv6 QoS Measurements Jordi Palet Martinez,
ConsulIntel, Spain

Comments:

Q: What is the scalability on 'plain vanilla' hardware? How many flows
can be monitored?

A: There hasn't been extensive testing and analysis on these
aspects. Sadly, we don't have a system for generating flows. 40-50 one
megabit flows shouldn't be a problem though.

Q: Have you compared your project with commercial platforms?

A: There weren't any when we started. We would like to work with these
new vendors though. Furthermore, we have been very active within the
standarisation process through the IETF; there are several drafts
published.

Q: Is your platform able to use a offline trace gatherer that can be
replayed, and then do calculations based on these historical captures?

A: This is something that is being considered for future iterations.

- 12:28:20 -

Z. AOB

Lastly, before closing the session, the chair delivered some final
thoughts where he dwelled upon the role of the working group.
Participants were urged to take part actively in its success. One of
the themes mentioned throughout the talks, particularly by Gianluca
Iannaccone, was "the need for a trusted third-party Measurement
Authority" for diagnostics, performance and research. The chair
believes the RIPE NCC could play this role.

Meeting closed.

- 12:29:28 -