[Babel-users] Some idea on cost calculation based on hop count, loss and RTT

[Benda Xu]

Hi StarBrilliant,

StarBrilliant <b13253 at gmail.com> writes:

> In case I didn't explain clearly in my previous mail, I think the
> current cost function is not suitable for long-range links:
>
> current_cost = hop/(1-loss)^2 + RTT
>
> I want to suggest a new cost function that may be better:
>
> my_new_cost = (hop+RTT) * sqrt(clip(loss))
>
> I wish we could have more flexible configurations for balancing
> tunneled networks.

This is a very interesting idea.  Do you have plans to develop a set of
test and benchmark criteria to objectively assess the effectiveness of
different cost functions?

I have got a feeling that it is hard to argue on which is better without
a performance test.

Yours,
Benda