Mesh networking: olsrd? b.a.t.m.a.n.? OpenWRT?

[Joshua Judson Rosen]

On 2015-02-25 14:33, maddog at li.org wrote:
> I have not been keeping up with what you have been doing, but if your target is moving, perhaps
> what you want is Mobile IPv6 instead of "mesh", or a combination of the two.

At a high level, I want to set up have some stationary servers, and a bunch of
mobile nodes running around at a brisk pace, with a minimum of cabling and
without having to deal with problems from RF shadows as the mobile nodes move
around RF-opaque obstacles. The abundance of opportunities for RF shadows in the
target environment means that can't just stick one Wi-Fi AP with a good antenna
in a high place and be done with it, and wanting to avoid cabling pushes me
toward (AFAIK) either mesh or WDS.

I'm kind-of biased against IPv6 right now--partly because I don't have much
understanding of it, and partly because so much of the equipment I'm using
just doesn't support it (though my OpenWrt Wi-Fi routers do...).

Somewhere in here I gather that I should (maybe) also be looking into
802.11r re: high-speed hand-offs.....



> Also, Andy Stewart, the former long-time head of the WPI Linux Group, and now
> the lead of the Chelmsford Linux Meetup group gave a talk on Mesh Networking
> at the WPI group last month.  I was not able to attend due to travel, but
> Andy is also into Ham Radio and Linux, so he may have been working on Mesh
> for some time.  I have copied him on this email.

Cool--hi, Andy.

> I would love to learn more about Mesh, so if you guys were willing to put on
> a presentation and/or workshop after you get things sorted out, I would be
> happy to bring some RPis to it so we could experiment.

Sure--why not. :)


-- 
"Don't be afraid to ask (λf.((λx.xx) (λr.f(rr))))."


> ----- Original Message -----
> Didn't the One Laptop Per Child project do mesh?
>
> On Wed, Feb 25, 2015 at 1:17 PM, Joshua Judson Rosen < rozzin at hackerposse.com > wrote:
>
>
> On 2015-02-20 09:17, Curt Howland wrote:
>> On Thu, Feb 19, 2015 at 5:20 PM, Patrick Flaherty < pflaherty at wsi.com > wrote:
>>> If you do get it working, it would be a great talk at a meeting.
>>
>> Agreed.
>>
>> Mesh networking is interesting, but the implementations appear
>> difficult to impossible at best.
>
> That... may actually not be the case anymore....
>
> When I posed the initial question, I was really sort-of grasping
> for any leads; but now I've learned enough to at least identify
> the different options, pick one, and even get one working.
>
> It looks like the prime contenders for mesh mechanisms are
> (roughly ordered in accord with the evolutionary timeline):
>
> - layer-3 OLSR mesh via olsrd
> - layer-3 B.A.T.M.A.N. mesh via batmand
> - layer-2 B.A.T.M.A.N. mesh via batman-adv
> - layer-1(!) mesh via 802.11s
> - layer-3 B.A.T.M.A.N. mesh via bmx
>
>
> The 802.11s mesh turns out to be remarkably easy to get up and running,
> following the HOWTO provided by open80211s project:
>
> https://github.com/o11s/open80211s/wiki/HOWTO
>
> (how well it works is yet to be seen)
>
>
> batman-adv appears to be more generalised than 802.11s:
> batman-adv can be used to aggregate any collection
> of layer-2 interfaces--including Wi-Fi (in infrastructure mode,
> ad-hoc mode, or any other mode), wired ethernet, PPP links,
> VPN links--where 802.11s (of course) can is usable only with
> 802.11 links (and then only with some chipsets).
>
> Presumably 802.11s and batman-adv are the most transparent
> options, since the other (layer-3) options rely on rearranging
> the *IP* route tables....
>
> It's still not yet obvious to me what to do about mobile nodes
> moving between the mesh nodes at speed--i.e., just how quickly
> the mesh can deal with the topology changing (or what knobs are
> available for tuning that), or if it makes sense to include
> traditional APs in the mix so that the roaming nodes just do
> traditional client dissociation/association cycles, or how
> to handle roaming nodes that aren't equipped to do mesh networking
> themselves.....