[WLANware] DHCP servers and Subnets in Wireless Mesh Networks [Mesh Architecture]

Tue Nov 6 16:32:34 CET 2012

On Mon, Nov 5, 2012 at 7:00 PM, Juergen Neumann <j.neumann at junes.eu> wrote:

> Hi Naman,
>
> > I am Naman Muley. I am involved with building a wireless community mesh
> > network in India.
>
> That's great! What kind of community network are you planning to deploy?
> To what degree do you want to include the community in building the
> network?
>
> For now, we want to start small. I wanted to deploy a mesh network inside
my undergraduate college campus. It is not so much a 'community' wireless
network. For starters, the involvement of the community at large is
minimal. Me and a group of undergrads are working on establishing it for
the campus students for now. It will also act as a good research test bed
for wireless research. But my major aim is to provide seamless networking
inside the campus.

Wiki: http://damndaiict.sourceforge.net/ (it's a bit outdated. Yet reflects
the current deployment)

>  > I wanted to understand the options for deploying DHCP
> > servers in wireless mesh networks. I wanted your opinions as the freifunk
> > community is highly active and experienced with mesh deployments.
>
> Yes. We are building distributed networks. We are also distributing the
> ownership of the network. So that the people own their own and maintain
> their routers. The DHCP servers are comletely decentralized.
>
>
> I understand that mesh nodes are all layer 3 routers. Generally, in a
> wired
> > network, we have one single DHCP server for the whole network and DHCP
> > relay agents at various routers down the heirarchy.
>
> That does not sound like a distributed network, but rather like a top
> down aproach. If you want a centralized service like a single DHCP
> server, you should focus on batman-advanced. This is a layer 2 aproach
> which is ideal for such an infrastructure.
>
> I was considering both the options and I wanted your opinions on both the
possibilities. To give a better idea about the scale of the network, the
area we intend to cover is roughly around 50 k sq mt, should support around
1500-2000 hosts. It's a campus basically. There is a lot of greenery, so
we're going to have to work out the positioning of the routers
strategically.

I had a look at batman-adv. It does seem to be ideal.
Yet, I had some conceptual doubts about distributing the DHCP servers as
opposed to having one DHCP server give out IP addresses from a large
subnet.

> > 1. Should a wireless mesh network big enough to cover a campus (for
> > starters) have one huge subnet or multiple subnets
>
> This highly depends on your social model. If you are planning a top-down
> network network design, then you should go for one big subnet. But if
> you are planning for a buttom-up distributed peer-to-peer network, then
> you should think of multible subnets.
>
>
Could you explain a bit further what you mean by a bottom-up distributed
peer-to-peer network? I had posted in this mailing-list earlier and Marek
had sent me a link to the webportal for Freifunk Slovenia. (I lost that
bookmark, I'd be glad to  have it again! :S ) There I remember freifunk
uses IP registry mechanisms. So anyone can come up with a router, register
certain IP addresses with the central registry and have the wireless
network deployed.

> > 2. A central DHCP vs multiple DHCP's assigning addresses from different
> > subnets?
>
> See 1.
>
> > 3. If we do have multiple subnets, how is the mobility affected? I mean,
> > because once you get an IP address from a subnet, you get a default
> gateway
> > in that subnet also. Once you move out of the range of that subnet's
> > router(s), how do hosts maintain connectivity?
>
> I your end-users are able to run the mesh-software (OLSR or batman) on
> their devices, then this is no problem, because they can choose a fixed
> gateway with both applications. If they are just "stupid" dhcp clients,
> then once again batman-adavanced as a layer 2 protocol would be the
> reasonable choice.
>
> So we don't deal with IP handovers at all? Also, here's a major doubt that
i had. For technically a mesh network to exist, do all the client nodes
have to be mesh nodes / nodes-with-OLSRd ? But I guess, mesh networks can
also be used by ISPs to run a network of routers giving services to clients
right?

> 4. In case we use IP handoff for the above problems, I'd be very greateful
> > if you could point me to the implementations for the same.
>
> See 3.
>
> > 5. How do you do it in Freifunk? If you have some ip registry links or
> > architecture links, I'd really really appreciate it!
>
> We are running quite chaotic networks. Our focus is not so much on
> seamless handover, but to provide maximum freedom and resposibily to
> every single individual node-owner. We want to run distributed networks
> to the maximum degree that is possible. Some things will not work in
> these kinds of networks, such as seamless handover to dhcp clients. But
> that is not our main focus. We just want to be as decentralized as
> possible.
>

So, each node is in itself a router? I mean.. I wanted to include general
users with cellphones. I don't know how to qualify the network I intend
here. I do want to provide VOIP service to users, dumb users. I mean, in
the future, I would really like independent people to come and include
their home-routers in the network, but I also want to support dumb users
with cellphones. I want to have telephony inside the network. I know this
is screwing with a lot of principles. So to get the picture clear, there
will be a network of MR's (mostly linksys54GLs ). There will be dumb users
who intend to obtain services of the network for telephony / VOIP calling
other fellow users inside the network.

> Thank you so much for all the efforts! I appreciate any help from the
> > community!
>
> What exactly are you planning? Are you heading for a infrastructure you
> are planning for the community, or do you want to include the community
> in setting up and building this infrastructure?
>
> Both! It's not an infrastructure network in the sense, I am not using any
switches / routers / dumb WAPs. Pure mesh network consisting of
Linksys54GLs. But the user community need not be independent people with
their own routers.

Say around 14-15 routers cover the area. As the crowd distribution is not
uniform, we might have direct point-to-point links if we find it's
impractical to cover the whole area with routers. But following are some of
the goals we are looking from the mesh.

(*Users - random people with cellphones / laptops. ) *
1. Sealmess networking - Users connecting to MR's must have all time
connectivity and any technical handovers must happen oblivious to the user.

2. Telephony - The network must support voip calling to other users inside
the network.

3. Research test bed - The network of MR's will provide an active testbed
for researchers.

4. Extending it out to the community - Once we have a proof of concept, we
would like to extend the concept into the city. (this is when I would
intend to become exactly like Freifunk networks. )

Are these good goals to have for a mesh network? I understand that points 1
and 2 are typical goals for an infrastructure network also. If you could
shed some light on the confusion here, it will be very helpful.

 Greets from Berlin!
>
> JuergeN
>
>
>
> Thanks a mega-ton!! I appreciate all the efforts and time you all put in
for my queries!

Naman

> _______________________________________________
> WLANware mailing list
> WLANware at freifunk.net
> Abonnement abbestellen? -> https://freifunk.net/mailman/listinfo/wlanware
>
> Weitere Infos zu den freifunk.net Mailinglisten und zur An- und Abmeldung
> unter http://freifunk.net/mailinglisten
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.freifunk.net/pipermail/wlanware-freifunk.net/attachments/20121106/5e91bd20/attachment.html>