[e2e] Revisiting RON ("traffic engineering considered harmful" thread)

[David G. Andersen]

Some more blatant project promotion about the RON project
follows:

In June, I chimed in on the "traffic engineering considered harmful"
thread with a pointer to the Resilient Overlay Networks project:

  http://nms.lcs.mit.edu/ron/

As a refresher:  RON attempts to route around down, or terribly
performing, Internet links by sending data through a friendly 
computer located in some other autonomous system, in a manner
inspired by the observations made in the Detour project.  If,
for instance, the link between MIT and Berkeley is down,
it might try sending my packets first to my home machine 
on MediaOne, and from there to Berkeley, to avoid whatever
bugginess was eating packets on the Internet2 connection.

We found (in actual experiments where we let RON predict
paths) that this kind of indirect routing was able to reduce
the number of observed Internet outages by a factor of
three to ten, and produced some nice improvements in latency
and loss as well.

The final version of the SOSP paper is now on our webpage,
and we've just released the datasets we collected to peek at 
its performance.  I figured the data might be useful outside
of the context of just RON;  there are a few days worth
of RTT and one-way loss samples between 12 and 16 nodes
that may be interesting for data mining purposes.

We're continuing to collect data from our testbed;  if
you find this kind of data useful, and have suggestions about
ways we could make it more useful to you or easier to
deal with, please let me know.  And, of course, comments
on the research or the paper are very, very welcome.

   -Dave

--
work: dga at lcs.mit.edu                          me:  dga at pobox.com
      MIT Laboratory for Computer Science           http://www.angio.net/