[Planetlab-users] PlanetLab nodes - serving ISPs and node interconnections

[Rex Chen]

Dear PlanetLab Community,

We are doing a research project using PlanetLab are as part of the
experiment, we would like to understand the ISP and interconnections between
each and every PlanetLab nodes. Although the Internet at large has complex
network interconnections,  I was wondering if some accurate statistics may
be available for PlanetLab nodes, perhaps via the PlanetLab administrators
or some monitoring tools.  For example, out of the 600+ PlanetLab nodes:

- 60% connected to a high-speed academic/research network (further
breakdown,  30% on ISP x, 20% on ISP y, 10% on ISP z)
- 30% connected to a tier-1 ISP (further breakdown....)
- 10% connected to a link with tier-2 ISP (further breakdown....)

- 50% of PlanetLab nodes are 10 or less hops away
- 40% of PlanetLab nodes are 20 or less hops away
- 10% of PlanetLab nodes more than 20 hops away
(I just made up all the numbers above)

Using a script to run traceroute from one PlanetLab node to all other
PlanetLab nodes, I found the following ISPs traversed by PlanetLab nodes:
     a.  Academic/Research network: cenic, ucaid (abilene), geant2 (Europe),
Pacific Wave, canet4 (Canada), cernet2 (China), APAN (Asia-Pacific), ja (UK)
     b.  1st tier ISP: Level3.net, att.net, sprintlink.net, qwest, Verizon,
ntt
     c.  2nd tier ISP and Others: cogentco.com, above.net, singtel, France
Telecom, Brazil Telecom, Time Warner Telecom, SBC global

In general, it seems majority of the PlanetLab nodes belong to "a."  since
most of the nodes are served by university institutions.  But there are some
country, continent, regional ISPs as well.  Simply using traceroute output
to deduce the numbers I made up doesn't seem very accurate for the
measurement study.

Any information, ideas for digging deeper would be very appreciative.

Thanks,
Rex

University of California, Irvine
Graduate Student Researcher
rex at ics.uci.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.planet-lab.org/pipermail/users/attachments/20060905/085aa180/attachment.html