RIPE 38

Draft minutes Test Traffic Working Group at RIPE 38


Chair Keith Godber
Scribe Rene Wilhelm

49 attendees

- -----------------------------

Agenda:

1.Agenda Bashing

2.Introduction of new chair

3.Status and plans, presentation by Henk Uijterwaal

4.Data analysis, presentation Daniel Karrenberg

5.GPS Site Survey tool, presentation by Florian Gruber

6.Competition

7.AOB

- -----------------------------

1. Agenda Bashing

Agenda adopted without modification.


2. Introduction of new chair

Matthew Robinson stepped down after the last meeting.
Following a call to the tt-wg mailinglist, Keith Godber
volunteered to take up this role.

3. Status and plans, presentation by Henk Uijterwaal

[slides at http://www.ripe.net/test-traffic/Talks/0101_RIPE38_HU/ ]

Henk presented status of and future plans with Test Traffic Measurements

TTM is now available as a membership service from RIPE NCC. As the tax
status is still under investigation, the service fee is currently set to
EUR 0.0; hardware is provided against cost. Well defined analysis and
research topics for next months.


Questions & Answers:


Q. Have you thought about adding a service like 'expected path'?
Measure the path taken by packets between two boxes, notify user
when it changes from what is expected. Would need a method to
put the expected path in a database.
A. We do have a database of traceroutes between all boxes.
Good idea to raise alarm if path changes from expected one.


Q. Testboxes rely on GPS, do you have alerting signal when quality is
bad?
A. Yes, in the daily plots we color code the status of sender's and
receiver's clock. We also have graphs and e-mailed lists of
satellite reception conditions.


Q. Can a test-box be used in a site's NTP service? would one have
to contact RIPE NCC to enable it?
A. Yes, a test-box may be used as stratum one time server in a site's
network; no need to contact us, the service is available from the
box as shipped. However, we do ask you to use it in a sensible way:
only a limited number of machines (up to ~10) should peer directly
with the test-box' NTP server; the rest of a site's machines should
take time from these secondaries. In future we may need to put
passwords on the NTP service, secure it with encryption and enforce
a maximum on the number of clients.


Q. Do you have access lists to get to a box' data?
A. Not at the moment, we want to add a service where people can get
to latest results of a box in near real-time via http; that will
require access control. All data are stored in a central database,
we can configure from there, athough at this moment there's not
much to configure (on request we provide data files via ftp).


Q. Is there any process for analyzing the data we have? to learn
traffic measurement variables?
A. You can get the raw data from us, in ROOT format [root.cern.ch];
instructions and first steps with ROOT in the framework of TTM
are posted on our website. Upon request we put the data on our
ftp site. If you have analysis code you find useful, we can discuss
it in the test-traffic working group and include it the daily TTM
analysis job.


Q. Do you have experience with Denial of Service attacks to test-boxes?
A. No.
Q. Are you prepared?
A. If we see it we react appropriately. Not many running services
on the test-boxes.
Q. But what about ICMP or UDP storms? These could disable a box.
A. Yes, but one would see these on more of the site's network(s),
not only the test-box would be affected.
R. Still your naming scheme makes it relatively easy for hackers
to guess a test-box name; suggest to switch to a naming scheme
that hides the purpose (TT measurements) of the box.


Q. Do you have a map which shows location of test-boxes?
A. Yes, available from the member pages.
Q. As antenna is outside, what about weather conditions? ice, lightning?
A. Snow and rain do not affect antenna itself; it comes as a sealed unit
which can be submerged in water. The connector, however, might get
damaged, as it is accessible on the outside. Usually, the problem
there will be liquids which could cause short circuits; easily solved by
disconnecting the connecting plug and letting it dry.

Lightning: any lightning protection will help, do not make antenna
highest point on the building. Note that it is the host's
responsibility to make it compliant with the local lightning protection
scheme. As each country has its own regulations in this area, it is
very hard to do coordinate that from Amsterdam.


4. Data analysis, presentation Daniel Karrenberg

[slides at http://www.ripe.net/test-traffic/Talks/0101_RIPE38_DK/ ]

Daniel presented some of his findings from looking at the data
over the last two weeks. The slides he showed illustrate what
interesting effects TTM can measure, and how the new plots-on-demand
service can help zooming in to particular areas of interest.

The precision the TTM tool can achieve was illustrated in the last
slide, which showed a change pattern with amplitude of ~200 microseconds
for delays between RIPE NCC office and Amsterdam Internet Exchange;
these could be traced back to load balancing over the two different
circuits that connect the two sites.


Questions & Answers

Q. Can you report the number of changes in traceroutes as an
indication of routing stability? would be especially useful
to know when different paths have same number of hops.
A. Yes, good idea, we'll implement this.


5. GPS Site Survey tool, presentation by Florian Gruber

[slides at http://www.ripe.net/test-traffic/Talks/0101_RIPE38_FG/ ]

Florian talked about his work on a GPS site survey tool, which
aims at assisting a test-box host in finding a good location for
the GPS antenna before starting hard installation work. He also
presented his experiences with various antenna positions.


Questions & Answers

Q. Windows sometimes work?
A. Yes; if you don't have a possibility to put the antenna on the roof
but do have a window facing south, in a high position it could work.
However, you would have to monitor it for a longer time; the GPS
satellites are not stationary with respect to earth, so good
conditions at one moment could evolve to bad conditions some hours
later.

Visibility of higher elevations is very important. On the outside
of a window (wall mount) is much better than inside. Finally,
the window will attenuate the signal; if coated with metal or
other reflective material, it could even go to zero.

General rule of thumb: if you can't get to the roof but can mount
the antenna on a southeast[*] wall, looking up, it will work.

([*] assuming you're on the northern hemisphere)


6. Competition

Only few entries were received for the 'renaming the project'
competition.
No real winning catchy name, so it was agreed to stick with the old
name of 'Test Traffic Measurements' and close this issue.


7. AOB

No other business; meeting closed.