Hi,

I'm looking for a cheap solution for building quite big (currently 50-100 users) wireless network for providing Internet access.

I don't want to use simple WDS bridges because of performance reasons. It seems that a mesh network would be best for me. I've googled a bit for mesh network routers, and found some products, but I'm interested in the Meshcube the most. The only reason I haven't already bought it is the price .

I would like to deploy "Scenario 2" described on Meshcube wiki (http://www.meshcube.org/meshwiki/HowTo_ … nScenarios) - one WiFi interface "works" as a mesh router and other as a standard 802.11b AP. So here is my question: can Linksys WRT54G with OpenWRT and required packages installed be used as described above? Does anybody use such network? Or maybe anybody succeeded in installing Meshcube's software on WRT54G? Any other ideas for creating a city-wide wireless network?

Thanks in advance!

Welcome to the quest for the holy grail.

I will begin deploying a 70-node mesh network this afternoon, after about 9 months of planning, coding, testing, and hair-pulling.

My network is based on the WRT-54G running openwrt and a set of custom packages based on open-source software.  Preliminary (i.e. "works for me") packages can be found at http://www.wildcatwireless.net/wrt54g
The packages are wildcatmesh, meshctl, and meshweb.  meshweb is the least complete of them all, providing a rudimentary web interface into controlling the network.  I intend to flesh it out over the next few weeks, but since I'm the only one doing network installs and maintenance, and I don't need a web interface, it's lowest on the food chain.

I have chosen to use OLSR as my mesh protocol (www.olsr.org) and secure the network using openvpn.  Obviously, there are some exciting advantages to using a smart router at the CPE, including edge bandwidth control and using the VLAN capability of the WRT54 to put 4 separate customers on one radio.
  I'll keep the community posted on the success and failures of a large scale mesh based on consumer grade equipment and open source software.  If you've got any coding skills, your help is always appreciated.

John Gorkos
Wildcat Wireless

The more I think about it the more I realize how tough it is - yes, you're right, for sure .

9 months?! You are patient. I've been playing with it for about 2 weeks.

After reading a bit more about WRT-54G I realize that you must have been using two connected WRT-54Gs, one of which works as an access point and the second one as a "mesh router" (ad-hoc client). Am I right? I'm asking because I haven't got any WRT-54G, so currently I can't play with it.

I prefer shell console to web interface, too .

This is exactly what I was thinking about.

Sure I have . I will definitely contact you as soon as I get some WRT-54Gs to play with. Thanks a lot.

Cheers!

Pawel Foremski

Er, no.  The WRT54 can be seperated into up to 6 seperate physical interfaces, one for each ethernet port on the back and one for the wireless network.  My clients aren't wireless, they're just inhabitants of an apartment building with an RJ45 jack with a note that says "Plug here for high speed internet".  Each building has 4 apartments, so I use the 4 lan plugs for 4 seperate VLANs, and that way the apartments each think they've got their own network.  The wireless radio is the mesh network node, and certain routers have a broadband connection on the WAN port.  If the mesh software detects an existing default route with a metric of 0, it advertises itself as a gateway node and passes traffic out to the internet.  One piece of hardware is the firewall router for the end user, the mesh node for the wireless network, and the gateway in certain situations.

John Gorkos

I see... I'll have to deploy a bit different solution, but similar (wireless clients). There should be no additional problems apart of radio channels; I'll have to think about it.

Coming from my other topic about a mesh-network I see some similarities in this topic.

I already have two WRT54GS running with OpenWRT. I wanted to let them do adhoc networking, which works for testing between one WRT and my Powerbook.

ab0oo: How do you connect the meshes? Is there always one client and one server? I think of one more or less self-constructing network, where routers find new ones and simply connect to them if these are also meshes. Can you connect more than two meshes in one adhoc network, or do you have to build connections for every route one by one?

It would be great if one can connect to such a mesh in a AP-like fashion, perhaps even via adhoc!? If I can jump on the mesh network and surf from my garden...

Perhaps you could post some details about your setup. I would realy appreciate it.

I'm doing HTML and a bit PHP, perhaps I could also help in the web interface part.

Best regards!

I'm not sure I get the "connect the meshes" part.  For me, there is only one mesh.  Lots of radios, but only one mesh.

My mesh network is called WildcatMesh, since my ISP is Wildcat Wireless Internet.  I have a series of packages that I load onto a router after I flash openwrt onto it.  After I load these packages, all I need to do is give the router a new static address and the address of the VPN takeout point, and throw it out into the network.  It will discover how to build out the mesh and how to route through its neighbors to the internet gateway.

  The kicker for most folks is this:  you can't use a meshed router as both an access point and a mesh router.   If you have your average windows laptop with a wireless card, and a couple of mesh routers, it takes a fair amount of work on the windows side to use the mesh network. 

Fortunately, Andreas Tønnesen (www.olsr.org) is getting ready to release windows binaries for his olsr implementation.  This will allow a windows machine to join up with an existing olsr mesh and participate as a node, giving the laptop user the ability to roam throughout the mesh and onto the internet.

  I'm working on a document that describes in detail how my meshing system works (it's a combination of openwrt, OLSR, and openvpn) to provide secure, bandwidth controlled internet access to subscribers in a pay-for-use environment (i.e. a wireless internet provider in an apartment complex).  I'll have it up on the Wiki in a few days.

John Gorkos

dear pjf

i have a cube, and for professional usings it is perfekt. the price is ok, (more ram, flash, cpu-power, up to 8(!) wlandevices )

no, for this scenario you need at least 2 wlan-devices.

yes at home in my living-room, and in the garden.

i haven't try that, but i think it's to less storage on the wrt54g, and i don't know if you need other drivers for the nic's.

maybe wrt54g is the cheapes way to build a mesh like you want, but imho not professional.
for homeusers the wrt54g is quiet good enough

greetings madd

Ok, I'll bite.

WHY is the WRT54G not "professional" enough for commercial installations?  It is not "carrier grade" equipment, but neither is the meshcube.  The WRT56G board is a solidly designed single-board computer with a remarkable amount of power for low cost.  It is environmentally challenged in that it is not designed with mil-spec components, but IMHO neither is the meshcube.  It is not highly resistant to lightning strikes, but again, neither is the meshcube.  By adding an external lightning arrestor to the antenna jack, you have a lightning-resistant device.
  I have WRT54Gs mounted inside RooTennas, mounted inside NMEA cases at the 200 foot level of towers, mounted bare naked in peoples' attics, and  sitting on my desk in the lovely blue case that Linksys provides, all of them providing excellent service levels.

So... what makes the meshcube "professional" versus openWRT on a WRT54G?  Price?

John Gorkos

oh, hmm i don't want to step on your feed, sorry.
- it's not the price, its more upgradetable, there are more diffrent wlan-devices availible (diffrent powers up to 200mw, of course diff. modes),
- and you have a bigger cpu. for example; you need cpu-power if you have some or many vpn's on a node. you cant have many vpn's on a wrt54g, and hold troughput.
- you have more space on board to install software you want/need without plug on a usb-storage.

madd

> i have a cube, and for professional usings it is perfekt.
> the price is ok, (more ram, flash, cpu-power, up to 8(!)
> wlandevices )

Currently the price is essential for me .

>> I would like to deploy "Scenario 2" described on Meshcube
>> wiki (...) - one WiFi interface "works" as a mesh router
>> and other as a standard 802.11b AP. So here is my
>> question: can Linksys WRT54G with OpenWRT and required
>> packages installed be used as described above?
>
> no, for this scenario you need at least 2 wlan-devices.

Yes, I know it now, I was misguided by two antennas .

>> Any other ideas for creating a city-wide wireless
>> network?
> maybe wrt54g is the cheapes way to build a mesh like you
> want, but imho not professional.
> for homeusers the wrt54g is quiet good enough

Of course, you can always use

http://www.radionet.com/265571.shtml
http://www.smartbridges.com/web/products/ah.asp
http://www.infinetwireless.com/Wireless … ument_view

...if you can afford them . As I said the cost of this network is currently important.

By the way, recently I've read tons of wireless mesh networks documents and have found many doubts about their scalability. I wonder is it possible to build a mesh networks consisting of houndreds of nodes with omni antennas.

This is why currently I'm considering building my network's backbone of several WRT-54Gs connected with directional antennas, and then connecting other WRT-54Gs to them (using 100BaseT) working as APs to provide client access. Of course I'll use traffic shaping, filtering and policing to block trash and possibly minimalize hidden node problem. Oh, and in case you were interested - here you can find tons of practical information on hidden node problem:

http://slashdot.org/articles/03/08/04/1416200.shtml

Still searching for the holy grail...

P.S. BTW: I hate all those f****** phpBB forums about serious technical topics - why there are no mailing lists? Do you need all those "emoticons", "BBCode", etc.?

It seems to me that a cheap solution to the two-radio problem would be to buy two WRT54Gs (at $54.00 each, not a bad investment.  See pricegrabber.com), or one WRT54G and one WET-11.  Use a little creative coding to set one wifi radio in AP mode, and put the other one in bridge mode, using it just for it's radio, as the mesh radio.  Then make the one running in AP more do all the computational heavy lifting.

It would probably work best using a WET-11 hooked to the WAN port of a WRT54, and using that as the mesh radio, and then using the built-in radio in the WRT54 as your AP.  The WET-11 works as a layer 2 bridge out of the box.  If you've got a little more cash, use Senao CB3+ bridges instead of WET-11s.  They're more reliable and do 200mw out.

So, you'd wind up with a WRT54g with 4 lan ports and an AP bridged together in a local network, plus a wireless bridge connected to the WAN port, with the WRT54G acting as the mesh router brains.  Should take about $120-$130/node and a few hours to set up.  The bonus is that you could seperate the mesh radio from the router by up to 100m of ethernet cable.  So you can put the AP in the basement, and the wireless bridge doing the meshing in the attic, power the bridge with POE, and have a great little setup.  For the cost of one "professional-grade" meshcube, you could have a two-node system.  With the sacrifice of a few brain cells, you could VLAN off one of the LAN ports on one of the WRT54Gs and connect it to your broadband connection (cable, dsl, ete) and have a gateway node for your mesh network.

Just a thought.
John Gorkos

I'm not offended, just involved in the debate.

I won't argue here.  There's a company out there with a three-radio mesh system that a meshcube would work great for.  They use one radio as an AP for local clients, one radio for "downstream" mesh traffic, and one for "upstream" mesh traffic.  If you put a pair of 802.11a radios as your mesh radios, you could use the 3rd radio for a B/G access point, and have a sweet setup.

I do a lot of VPN over my WRT54G, and you're right:  the CPU can handle one tunnel at 2.5MB/s symetric.   But, since I throuput throttle my clients to (at max) 512kb/s each, even all 4 clients using their full bandwidth doesn't max out the CPU.  I do NOT use a WRT54 as the mesh take-out point, for this very reason.  I have to use "real" computers as VPN gateways.  But that's not necessarily bad, because I have lots of other things I want to do at the takeout point (NAT, firewall, management, etc), so I'm ok with that.

I've debated this in my head a hundred times.  The bottom line is:  I want my mesh routers to do ONE thing:  act as CPE for my ISP.  To do this, I need the meshing software, the bandwidth control software, and the VPN software.  If I had 32 MB of storage on the router, I'd be tempted to do stupid things like put perl on it and write tons of scripts that give me second-by-second statistics, etc, etc.  That's not what I want the CPE to do and it would take time and effort away from running my business.  By having limited space on the WRT, I'm forced to limit what it does to ONLY the things that it has to do to be useful to me.
  On the other hand, it would be pretty sweet to run a squid cache on every CPE box.
I guess the bottom line is that, for a smart, robust network, a mix of  a few higher level meshcubes and more lower level WRT54s at the customer site would work well.  If I could put a mini-PCI encryption accellerator in the meshcube, that would be a huge win, too.
  I'm so busy trying to deploy my mesh to paying customers, I haven't the time to play with new hardware, or I'd buy a cube to try it out.

John Gorkos

I'd rather use WRT54G as the mesh router - I want to fully control my backbone (ad-hoc mesh network). Such setup would also be more flexible - I can choose of many APs to connect to the router, or even use ethernet switches on some (cable-only) installations. What is more, WET-11s (about 130$) are more expensive than WRT-54Gs (about 100$) in Poland .

However, thanks for a suggestion. I'm still thinking about 802.11 mesh networks scalability. Anybody tested it?

Since I am working on a similar project I am really interested in:
1) the way you manage to control the access of your subscribers? I have a small WISP system using PPPoE and HotSpot . (authenticated via Radius) I want to implement all this in a mesh network.
2) Do you have a procedure in case you want to remotely upgrade the firmware of the WRTG at your subscribers site?