[Battlemesh] Multi-path over wireless

Juliusz Chroboczek jch at pps.univ-paris-diderot.fr
Sun Jun 19 08:50:38 UTC 2016


Hi Mitar, long time no hear.  Perhaps you can come to IETF Berlin this
summer so we can have a chat?

> I am thinking that I would like to try if there is better performance if
> instead of one 80 MHz wide WiFi point-to-point channel I would use 2
> radios with 40 MHz wide channels. The idea is that throughput would be
> the same, but because channels are narrower, there will be less noise
> and less packet loss.

This is a very interesting research project, but testing will be
difficult.  I think it could be a Ph.D. topic, or perhaps a very good M.Sc.

> The question is how to achieve this? The problem is that link speed
> fluctuates so round-robin multi-path seems that it will not work well
> over wireless.

There's more to it.  Traditionally, multipath is about building disjoint
routes.  With wireless, you need to pick non-interfering routes, which
means knowing about radio frequencies.  Even if you make the simplifying
hypothesis that disjoint frequencies don't interfere (see Baptiste's comment),
this requires having some extra information.

Now it turns out that Babel-Z carries all the information you'd need to do
that, and I'm going to have a student working on Babel-Z this summer.
However, the plan is to just do some experimental work with the existing
code, not to extend it to new applications.  But experimenting with
multipath is something that I'd be very keen to collaborate on.

-- Juliusz


More information about the Battlemesh mailing list