[CWN-Summit] Seeking advice on low-cost mesh node wifi in St. Louis, Missouri, USA

L. Aaron Kaplan aaron at lo-res.org
Tue Nov 18 19:53:22 CST 2008


After the last CWN summit I have been digging deeper and deeper into  
accton based routers (FONs, merakis) and up until very recently the  
wifi drivers had very serious problems.
This was a big step back for me. Especially after the announcement at  
the CWN.
I am now slowly getting to where I wanted to be with these devices  
but nevertheless ... for building a *stable* network , I can not  
fully recommend them anymore as the only solution.
I think they are great for home users / CPEs where an occasional  
reboot won't matter so much.
But for backhaul, I believe we need something else.

What does that mean?
Just getting a stock (closed source) Meraki or (open source) open- 
mesh.com accton router wont' *currently* scale up to large networks  
(driver issues - still problems with VAP + ahdemo mode in madwifi) .  
This is my latest knowledge of open-mesh.com's dev status. Me and  
Antonio ported over OLSR to open-mesh.com , so now it also runs OLSRd.
However I absolutely love the dashboard with open-mesh.com. I think  
something like that is the way to go. I might suggest looking into  
orange-mesh (a clone) or pushing Michael B. a bit to release the open- 
mesh.com dashboard as open source ( I was told he would like to have  
that anyway. The firmware is open source already)
So, open-mesh.com is a great asset.

Concerning management software: more rigidly programmed, but more ISP- 
centric,  you can also use the funkfeuer.at management software:
I personally would believe this has more of an enterprise/ large  
network touch. It has all kinds of hooks to trigger different update  
tasks all from the central DB.
It has plugs into a map service, makes sure that IPs are assigned  
properly, is used for registering locations and nodes and new members  
etc. It is probably very similar to what Ramon described for the  
guifi.net. Wolfgang can tell you more about it since he programmed  
it. It works great in Vienna!

For backbone nodes, I developed the "mangrovia" box with some  
friends. It basically contains an very good pcengines Alix board ( I  
am *very* happy with those. They are well worth the extra price).
They contain a CF slot. So what I did was I created an Image builder  
and my Makefiles will generate different images for each box  
automatically - preconfigured!
This way we installed a small muni wifi network in Piemonte (northern  
Italy) in a couple of days. (ok, it was a week but we were lazy and  
the food was so good).
We also used some very good 5GHz backbone links in this network.  
Highly recommended.
With this setup we again provide connectivity with one single uplink.
Key was solid components, solid engineering, careful assembly (HF  
cables!) and experience. Now the backbone runs on 108MBit (~ 50- 60  
MBit/sec netto).
Ah and BTW: yes, it "meshes" over everything - LAN cables, adhoc  
links, managed links.

I will continue working on the ultra cheap FONs given my tight time  
resources. But I believe by now that they will only be part of the  
whole puzzle (mere CPEs, edge nodes of the mesh). Madwifi  
unfortunately got only stable in adhoc mode recently. For backhaul I  
hope I will get my fingers dirty with pronghorn boards soon and some  
more/different ALix boards.

Questions to the readers:
Anyone experimented with 802.11n outdoor? Does it make sense in  
practice? AFAIK it was built for indoor purposes.

pic: mangrovia box in bella Italia

-------------- next part --------------
Skipped content of type multipart/appledouble-------------- next part --------------

I hope I could help with sharing my thoughts on the issue. I am sure  
others have good additions or corrections to the above said.


On Nov 17, 2008, at 9:22 PM, Ben West wrote:

> Hello all,
> I should first apologize if I am posing an oft-asked question, but I
> find myself at an impasse, even after 2+ years of casual research and
> spectatorship in the Mesh Node Wifi movement (not to mention twice
> attending the CWN conference).  The diversity of participants in Mesh
> Node Wifi is awesome, but it can make feasibility research difficult.
> I work/volunteer at an activist community center (CAMP, stlcamp.org)
> in south St. Louis, and a local foundation just put out an open call
> for proposals for investing a substantial sum into community
> revitalization projects in the neighborhood.
> These 2 articles about an $8500 deployment of Meraki devices along a
> 2mile corridor in Kentucky motivated me to pitch a similar idea for
> this St. Louis neighborhood:
> http://www.govtech.com/gt/377232?topic=117699
> http://www.wireless-nets.com/resources/tutorials/low- 
> cost_mesh_hotzone.html
> However, further research in Meraki has yielded some unpopular
> business decisions they made just this year:
> http://en.wikipedia.org/wiki/Meraki#Criticism
> http://hardware.slashdot.org/article.pl?sid=08/03/24/1318226
> http://www.dailywireless.org/2008/11/05/meraki-diy-munifi-for-10kmile/
> I certainly understand Meraki's motivation to protect their market,
> but my impression is that decisions to lock down hardware make their
> products less viable in areas where wifi groups may face direct
> competition from established ISPs.  (I.e. can't hack routers to
> support QoS or customized captive portals).  The latter is actually
> directly relevant to my proposal, since I'm aiming for a captive
> portal hosting local ads to provide some operating revenue.
> So, assuming you have a $10-$15k start-up budget (including equipment
> purchase. deployment, AND marketing) for installing wifi along a
> ~2mile corridor with lots of 3story rooftops, what suggestions are out
> there?
> Meraki, and take your lumps?
> Open-mesh.com, which is appealing since OpenWRT can be deployed to
> legacy devices like residents' existing Linksys routers?
> OpenWRT + Kamikaze + OLSRd (i.e. roll your own)?
> Freifunk.net?
> WifiDog for the captive portal + OpenWRT?
> The basic, 1st order requirements for the Mesh network are such:
> - Robust & stable (this will be a funded deployment, and sadly not a
> dev project)
> - Low-cost equipment (population density of this neighborhood makes
> antenna strength 2nd order)
> - Capacity for centralized admin console
> - MAC tracking and auth (i.e. how many unique wifi clients have  
> connected)
> - Quality of Service (we anticipate lots of folks trying to run file
> sharing, whether sanctioned or not)
> - Customizable captive portal
> - Ability to route to multiple DSL connections from different ISPs  
> w/in the mesh
> 2nd order requirements
> - Support for legacy routers (e.g. able to flash old Linksys products)
> - Good transceiver strength
> - Integration with PayPal-like subscription payments
> 3rd order requirements
> - Mechanism to control per MAC access based on # bytes downloaded,
> e.g. "We see you've downloaded 3GB this month w/o paying for your
> access..."  This would be a very appealing way to provide limited free
> access, i.e. make the service more competitive, but then enforce fair
> cost sharing in case folks opt for sustained freeloading.
> - Ability to dynamically divert sessions away from congested DSL
> uplinks.  (I hope that having multiple DSL connections in the mesh
> will give us composite reserve bandwidth we can actively allocate to
> handle sporadic traffic peaks.)  Do conventional Mesh Node
> implementations already support this?
> - Ability for wifi clients to connect to each other (Meraki does not
> support this)
> 4th order hopes and dreams
> - Support for integrating a centralized squid-like HTTP caching
> server.  I.e. commonly surfed traffic gets cached within the mesh.
> I consciously anticipate this mesh node deployment to be a temporary
> thing.  The goal is to establish a wifi-savvy neighborhood presence
> that can use its collective buying power in the next few years to
> transition to new technologies, White Space devices in particular.
> Any suggestions would be gladly welcome.
> -- 
> Ben West
> westbywest at gmail.com
> http://savetheinternet.org
> _______________________________________________
> CWN-Summit mailing list
> CWN-Summit at lists.cuwireless.net
> http://lists.chambana.net/cgi-bin/listinfo/cwn-summit

More information about the CWN-Summit mailing list