[Battlemesh] The 'blowing up the network' results

Toke Høiland-Jørgensen toke at toke.dk
Wed Aug 12 16:42:33 CEST 2015

Juliusz Chroboczek <jch at pps.univ-paris-diderot.fr> writes:

> Right, although I'm surprised it didn't go worse than that -- the
> network didn't collapse, it's just the latencies that shot up.

Yeah, I was a bit disappointed at that. But then we only ran the test
for 60 seconds. Might need longer to collapse completely...

> Can you explain why the download and upload tests yield different
> results?

I'd be wary of concluding too much about the behaviour from this one
test run. There is so much variance that I'm not sure I would conclude
that they are meaningfully different.

That being said, I have seen immensely asymmetric throughput in similar
tests before, but it depends on driver configurations which direction
tends to win.

> Are you already working with Matthieu in setting up a web page that
> explains all of the results in comprehensible terms, or shall we do
> that when you visit in Paris later this month?

Let's put that on the agenda for Paris.

> -- Juliusz (who is cleaning up his flat in anticipation of Toke's
> visit)

I'm sure it'll be immaculate... :)


More information about the Battlemesh mailing list