[Babel-users] ETX metric for bandwidth limited links

[Benjamin Henrion]

On Thu, May 20, 2010 at 6:25 PM, Juliusz Chroboczek
<Juliusz.Chroboczek at pps.jussieu.fr> wrote:
>>>> Does it support asymmetric links?
>
>>> Yes.
>
>> Maybe I have not asked the question correctly: does it selects (by
>> default) in both directions the best path (even if they are different)
>> or does it selects only one path which is the best for both
>> directions.
>
> The Babel algorithm choses paths of smallest metric.  This can, in
> general, lead to asymmetric routing.
>
> The two metrics implemented in the current version of babeld (2-3 and
> ETX) are symmetric -- i.e. the current implementation will give the same
> costs to links in both directions.  This implies that in a stable
> network, routing will be symmetric unless you manually tweak costs and
> metrics (using e.g. filtering).
>
>> Is there some reason why would be prefer in our networks paths which
>> are the same for both directions?
>
> Please read the ETX paper[1] to understand why ETX is defined the way it
> is defined.
>
> Babeld is fairly modular, and it's trivial to experiment with new
> metrics in Babel, so if you have an ideas about how to improve on ETX,
> I'm listening.  (But please read Sections 3.4.3, 3.5.2 and 3.6, as well
> as Appendix A of the Babel draft[2] before you make any suggestions.)
>
> --jch
>
> [1] De Couto, D., Aguayo, D., Bicket, J., and R. Morris, "A
>    high-throughput path metric for multi-hop wireless networks",
>    Proc. MobiCom 2003, 2003.

ETX does not take into account the physical characteristics of a path.

It relies on a statistic made at the IP layer counting the number of
packet loss.

It does not make any distinction on whether the link is a gigabit one or a 56k.

Correct me if I am wrong.

--
Benjamin Henrion <bhenrion at ffii.org>
FFII Brussels - +32-484-566109 - +32-2-4148403
"In July 2005, after several failed attempts to legalise software
patents in Europe, the patent establishment changed its strategy.
Instead of explicitly seeking to sanction the patentability of
software, they are now seeking to create a central European patent
court, which would establish and enforce patentability rules in their
favor, without any possibility of correction by competing courts or
democratically elected legislators."