> I have just taken for granted that hnetd/HNCP would have functionality to
> configure routing protocol metric per interface.
Babel is designed to cope with metrics that change continuously (on the
order of seconds, much faster than what HNCP is designed to deal with),
and it has a number of mechanisms that refrain from announcing small
metric changes -- intuitively, Babel will reconverge quickly if it loses
its selected route, but will delay announcements and reconvergence if the
current route is good enough, just not the best.
The reason that works is because Babel has a strong loop avoidance
mechanism -- the intuition is that if you're going in roughly the right
direction while guaranteeing loop-freedom, there's not much that can go
wrong. I believe that this is difficult to do in link state, but David
Lamparter has been doing impossible things lately, so I'll refrain from
making any stronger statements.
The current code implements three different metrics:
- Appendix A.2.2 of RFC 6126 defines a simple packet-loss metric (ETX);
- draft-chroboczek-babel-diversity-routing defines a metric that takes
radio interference into account;
- draft-jonglez-babel-rtt-extension defines a metric that is suitable
for intercontinental overlay networks.
Of course, all of these extensions interoperate with the base spec, with
each other, and with source-specific routing. And all of them are being
used in production.
The packet-loss metric is evaluated in [1]. The overlay network stuff is
evaluated in [2]. No formal evaluation has been done of the radio
interference metric, which is something I really need to do ASAP -- we're
the only routing protocol that does something like that as far as I know.
[1] M. Abolhasan, B. Hagelstein, J. C.-P. Wang. Real-world performance of
current proactive multi-hop mesh protocols. Asia-Pacific Conference on
Communication (APCC 2009), Shanghai, China. 2009.
http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1747&context=infopapers
[2] B. Jonglez, M. Boutier and J. Chroboczek. Delay-based routing.
Draft. http://arxiv.org/abs/1403.3488
If people are interested, there's a 50-minutes talk about Babel that I've
given before, we could grab a projector and have me speak about my
favourite subject.
-- Juliusz
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet
