[Battlemesh] Proposal for battlemesh v10 "testbed"

Henning Rogge hrogge at gmail.com
Thu Dec 15 09:07:52 UTC 2016


On Thu, Dec 15, 2016 at 7:29 AM, Linus Lüssing <linus.luessing at c0d3.blue> wrote:
> On Thu, Dec 15, 2016 at 12:42:09AM +0100, Juliusz Chroboczek wrote:
>> > Simon mentioned that he wasn't really personally interested in
>> > "battling" protocols so much (as in the first few battlemeshes),
>>
>> This would be the kiss of death for Battlemesh.
>>
>> Battlemesh is about experimentally comparing the performance of routing
>> protocols.  If Battlemesh doesn't compare routing protocols, it becomes
>> yet another community meet-up.  The technical people will no longer come,
>> the community people, who come to battlemesh because it's a chance to meet
>> the technical crowd, will go to places that are more interesting, and
>> Battlemesh will wither away.
>
> To add one more provocative point: The numbers from the BattleMesh
> so far were never really helpful for the BattleMesh from the
> batman-adv developer side. Maybe the ones from the BattleMesh with
> OLSRv2 and its throughput metric were, for the first time -
> incredible work from Henning! (and resulted in batman-adv
> implementing throughput metric in BATMAN V, too (*) - I know from a talk
> at WBMv4 six years ago, that BABEL is very flexible regarding metrics and
> that it is sooo easy to change - so I'm wondering, why was there no throughput
> metric for BABEL at the last BattleMesh, or did I just miss that?)
>
> The numbers which were far more helpful for me and lead me to some humble
> contributions to batman-adv were from actual, real world community
> setups. Some measurement, statistics and graphs helped to set the
> right priorities to scale batman-adv to not just seventy - that's
> where we hit our first scalability bottleneck with batman-adv - to
> over a thousand nodes in various setups these days.

I agree to this... Battlemesh is unique for me because its one of the
few platforms that is centered on the routing protocols, which are a
"do or fail" point in mesh networks...

if we talk about "extending" Battlemesh, extending it towards the
lower layers would be the right decision. There has been a lot of work
on the Wifi drivers and queuing disciplins in the last years, things
we built the routing on top (or which directly interact with the
routing). That why I liked Dave Taht's presentations that much.

We have still to learn A LOT about how our routing protocols and the
wireless stacks interact.

Extenting (or moving the focus of Battlemesh) towards the higher
layers (like whole firmwares) mudds the water for the technical part
of meshing a lot.

> So, all these years, was the BattleMesh a waste of time for the BATMAN
> developers? Will they disappear from the WBM? No, far from it! It
> has always provided us with:
>
> 1) Inspiration: Talking with great people lead to great, new ideas.
> 2) Collaboration: batman-adv devs live in various
>    places around the world, so WBM is always our annual
>    meetup, to talk and work on things which are easier
>    in person than over chat, email or phone. (And sorry for others
>    that batman-adv devs then here and then seclude themselves a bit :(.
>    But we usually arrive with a pile of topics regarding batman-adv
>    which we want to discuss and work on during this short week)
> 3) Motivation: Talking with awesome people, especially the ones
>    doing a lot of crazy stuff in their community setups and
>    sharing experiences, needs and anecdotes, is always motivating
>    (at least for me)
>
>
> Juliusz, I agree that trying to gather conclusive numbers is
> something we should strive for on the long run. There are enough
> wrong claims and statements in politics, economics or marketing
> surrounding us already :( (and choosing a hip protocol name and
> marketing it with T-shirts etc. probably adds to that at WBM,
> too, unfortunately, ... but so did generelized claims like
> "layer 2" does not scale or "link-state" does not scale).
>
> Just saying, I think mesh protocols are still at a stage too early
> to actually be comparable with each other. Even after ten years of
> WBM we are still at the beginning of the possibilities of mesh
> networks. The only thing we can and I think should do, is trying
> to create a welcome, inspiring and motivating environment to
> experiment with and discover those possibilities and exchange our
> findings.
>
> It does not matter who is quickest at crawling for
> 20m if we haven't even discovered running on two legs yet
> and if there is a whole marathon still ahead of us.
>
> Regards, Linus
>
>
> (*): Or maybe Antonio actually started working on a throughput
> metric in BATMAN V before that BattleMesh in Slovenia? Can't quite
> remember.

I think to remember that we talked a lot about throughput metrics on
the BattleMesh in Leipzig because I had an experimental version of my
throughput metric with me.

Henning


More information about the Battlemesh mailing list