RIPE 45

Test Traffic WG - RIPE45 - Amsterdam - 15 May 2003
--------------------------------------------------

Chair: Henk Uijterwaal (RIPE NCC)
Scribe: Mark Santcroos (RIPE NCC)

Agenda:
-------

A. Opening
B. Administrivia
* Appointment of Scribe
* Attendance Sheet
* Agenda bashing
* Minutes of last meeting
C. TTM Status and Plans - Henk Uijterwaal
D. RIPE NCC Hostcount in the 21st Century - DFK
E. Discussion

=-=-=-=-=-=-=-=-=-=-= break =-=-=-=-=-=-=-=-=-=-=-=-=-=

F. The Spectrum of Internet Performance - Roberto Percacci
G. First Measurements with IPv6 - Henk Uijterwaal
H. AOB

A. Opening
----------
Henk opens the meeting at 9:10 and welcomes all attendees.
He outlines the agenda and introduces the first speaker: himself.


B. Administrivia
----------------
Minutes of RIPE 44 approved.
He announces that the chair (Keith) can not make it anymore to most
meetings and that he has resigned. So the WG is looking for a new chair.


C. Test Traffic Measurements - Henk Uijterwaal
----------------------------------------------
Slides at:
http://www.ripe.net/ripe/meetings/ripe-45/presentations/ripe45-tt-status/

Questions: None


D. RIPE NCC Hostcount in the 21st Century - DFK
-----------------------------------------------

No slides, but see: http://www.ripe.net/ripe/docs/ripe-271.html

RIPE was already doing measurements before the NCC.
The current model is that a limited set of people cover for the costs of
the measurements. Due to not enough hosts this is not realistic.

Why are we doing measurements:
- Statistics about address usage and routing is important for policy
making.
- Measurements statistics for day to day operations
- Need for authoritative measurements by government and general public

Cross subsidizing

Strengths as RIPE NCC for doing these measurements:
- Neutrality
- Staff doing the measurements have experience in doing measurements
- High quality data
- Only active measurements -> no privacy issues
- Long term data -> nice for trends

Weaknesses:
- Output too detailed (Too much of an academic attitude)
- Closed user group service - only participants can see data

Proposal:

Build on our strength and continue with high quality measurements.
Continue to do active measurements and not look at other traffic.
More Incident Reports as done for the saphire worm and the attack on the
NCC office.
Develop more simple to use products that are immediately useful.
(But no traffic light kind of things)
As come out of the survey: do more things that are generally available.

Personalized view of measurements.

Discontinue TTM as a closed user group. Make data available to general
public.

So why are we still charging for it: having a probe in your network still
makes it more interesting and useful for you. But the charge will be
significantly less than it is now.

Structurally measure DNS (root) server quality. React on reports in
press about bad DNS.

Will do significant effort of making the products more usable. As the NCC
has a long history of doing training courses we will extent that to the
measurements. Provide better documentation.

Depending on the feedback we get we will approach you in the near
future about the direction we will go in.

Next step will be to go the membership and ask for permission to use a
certain amount of the budget for these measurements.

All of this is also in RIPE-271.


E. Discussion
--------------------------------------
Q: Is this your idea or is it already presented to the board and such?
A: It's being discussed inside the RIPE NCC and with the board.

Q: Would be interesting to investigate a setup without the necessity
for a GPS clock.
A: Yes

Q: We are academic: no problem. But there might be some carriers / ISP's
that wont want their data published.
A: We don't have negative responses yet, but we didn't officially ask
yet.

Q: How about half open/half closed?
A: That will not solve the problem for doing things for a small group.

Q: Maybe move to voluntary disclosure?
A: Again, that will not solve the problem for doing things for a small group.

Q: Will you offer all details?
A: Yes

=-=-=-=-=-=-=-=-=-=-= break =-=-=-=-=-=-=-=-=-=-=-=-=-=

F. The Spectrum of Internet Performance - Roberto Percacci
----------------------------------------------------------

Slides are avaialble at:
http://www.ripe.net/ripe/meetings/ripe-45/presentations/ripe45-tt-performance/

paper is available online at:

http://moat.nlanr.net/PAM2003/PAM2003papers/3693.pdf

More quantitive results instead of qualitive. This would be useful for
generating simulations and as a method of measuring the quality of the
connection of a host to the internet cloud.


Questions from Anja Feldman, U.Munich

Q: How much of your results depends on the placement of the host
A: Since we measure the ratio of RTT over distance, the single
measurements depend very much on placement. However, we expect that
statistically the resulting distributions do not depend much on
position.


Q: You completely ignored the time it takes to actually send the data
A: Hosts are at sites that are well connected

Q: What is the maximum range in terms of values that you can ever expect?
A: As you can see in the samples we see values of the variables up to
a few hundred but the statistics becomes quite poor. we expect it can
get larger too



G. First Measurements with IPv6 - Henk Uijterwaal presenting for Florian Frotzler
-------------------------------------------------
Slides at:

http://www.ripe.net/ripe/meetings/ripe-45/presentations/ripe45-tt-ipv6/

Questions:

Comment from Gert Doering on v6 being faster than v4 in some cases.
He said that it is because of non commercial transit for v6 which in some
cases is a directer path.


H. AOB
------
No other business.


K. Closing
----------
Henk thanks all speakers and attendees and closes the meeting.