You are here: Home > Participate > Join a Discussion > Mailman Archives

Re: [mat-wg] suggestions for atlas service

  • To: Piotr Strzyzewski <Piotr.Strzyzewski@localhost
  • From: Robert Kisteleki robert@localhost
  • Date: Fri, 13 May 2011 11:56:01 +0200
  • Cc: mat-wg@localhost
  • Organization: RIPE NCC


On 2011.05.05. 17:15, Piotr Strzyzewski wrote:
> Hi
> Two suggestions:
> 1. Some looking-glass from the probe. Simpe ping/traceroute from the
> probe triggered from the web interface could be nice.

I would think that this will be covered by the user defined measurements
feature, where the measurement is a one-off, not a continuous one.

> 2. Better probe location display. Currently I host two probes, both 10cm
> one from each other, put into two different ASes. When I have put
> exactly the same geo data, the "display" mechanism put one of it onto
> the other one.

This is a tricky one.

Background information: we'd rather not publish the exact geo-coordinates
the hosts put in, because of privacy concerns. So what we do now is that
whenever we display probes on a map, we round the geo-coordinates to two
digits. This introduces a "random" error of up to 1000 meters or so. (If you
zoom in on Amsterdam, where we have enough probes, you'll see that all
probes are aligned into a neat, sparse matrix; it's because of this.)

As a consequence, if two probes are located very close to each other, the
rounding makes them show up as overlapping. You could try to move your probe
a bit, like to the next street or so to see if that helps. The RTT map is
probably the best page to check the outcome.


> Piotr