[CUWiN-Dev] G and A and realworld throughput

[Stephen Ronan]

Stelios Valavanis wrote:

>i couldn't listen to the metrofi "radio" piece (what the hell is a 'wax' 
>file). from their site i see thye are using proprietary wireless with some 
>kind of CPE. i wish they had more on the backhaul (my issue here) which i bet 
>was in the "radio" piece.
>  
>
Here's info about .wax files
http://forum.dbpoweramp.com/showthread.php?t=4992

You can grab the mp3 files for each show here:
http://www.muppethouse.com/wirelesstechradio/

In particular this is the show with the MetroFi segment.
http://www.muppethouse.com/wirelesstechradio/WTRshow062905.mp3
The MetroFi segment starts at 19:00 minutes with the key part re: 
backhaul starting at 23:37

>how does roofnet compare with cuwin?
>  
>
I think generally speaking they're occupants of the same niche. Perhaps 
roofnet could be a little sooner to offer a very cheap mass market 
solution for situations where there are rarely if ever more than about 4 
hops from the wired gateway and perhaps cuwin is more ambitious in 
regard to larger scale deployments further down the road.... but I'm not 
expert in either. I've installed roofnet on some netgear routers and 
used it at home, have participated in a small and then a 40- node brief 
test with it installed on Soekris boxes, and am currently coordinating 
some testing of 20 of them at a mixed income housing development. I 
haven't installed and used cuwin at all yet. My general sense is that 
roofnet is pretty solid at this point on the Soekris boxes, and there 
have been some major kinks to work out with the netgear equipment, but 
it may not take much longer for that to be completed.

Up till now on the netgear wgt634u, roofnet software has been loaded on 
alongside much of the existing netgear software and the latter has been 
a moving target with netgear releasing one firmware version after 
another with unpredictable results on the roofnet part. The MIT 
researchers now plan to cease relying on the netgear software and 
replace that portion so that the entire package is more predictable. As 
of today there have been some continuing significant problems, but they 
found a major problem in the interaction with netgear software this 
evening, expect to be able to upgrade the set of 20 boxes we're working 
with as of tomorrow, and I'm hopeful that those will then become stable.

In regard to your plan, I'd continue to wonder whether fewer backhaul 
nodes with directional antennas that communicate with just a few other 
nodes might make sense.

Tangentially, will some of the Cisco equipment be 1300 series that could 
be used in bridging mode? If so, I'd wonder if some use of them in that 
regard could reduce the number of backhaul hops. But that might not be 
necessary or helpful.

In regard to the site, what you sent was helpful but I still have 
questions about what it will be like. Are you dealing with lots of 
difference in building height and trying to cover buildings from several 
angles or is it pretty much up and down one street with relatively 
uniform height, thin buildings on each side of the street that you're 
trying to cover? Will the nodes mostly or all be on roofs? Would be glad 
to see a map (and it might help me think out how one might deal with 
similar circumstances in Boston).

I can't answer your question regarding 802.11a distances and 12 dbi omni 
antennas and also would also be interested to know the answer. As you 
probably know Cisco 802.11a bridges are expected to have 2+ mile links 
in a point to multipoint setup with a sector antenna at the root and 
outlying high gain directionals. But that's a very different situation.

  - Steve Ronan

>here are some more details on what i'm building. i actually would not mind 
>sending a map to individuals but not to the list as it is for a client.
>
>there are 15 nodes using atheros 802.11a radios with cuwin as a backhaul. i 
>used omni but flat (donut shaped - don't know what you call it) antennas. 
>these nodes feed separate 802.11g APs with sectorized antennas meant to 
>penetrate into buildings across the street. the whole length of the site is 
>probably 2000 feet so i can probably see every node from every other but 
>since cuwin does not necessarily find the fewest hops to it's backhaul point, 
>i can't guess how it's going to perform here. it could go as many as 8 hops 
>(ouch!). i could do a few things to minimize this. i could use unidirectional 
>antennas to make sure i'm connecting to more distant nodes closer to the 
>uplink point. i could also add a second meshnode at the uplink point and 
>"break up" into 2 mesh networks but that doesn't reduce hops any more than 
>getting my more distant nodes to connect via nodes closer to the uplink. i 
>think uni antennas is a better shot. what kind of distance should i expect 
>with 802.11a atheros (100mw i think) and omnis (hypergain HG812U-PRO 12dBi) 
>anyway? is there anyway other than with unis to force meshnodes to connect 
>via particular nodes?
>
>
>On Friday 15 July 2005 08:42 pm, listsubs0506 wrote:
>  
>
>> >Stelios Valavanis wrote:
>>    
>>
>>>shall i assume that the cuwin is no different and that this would be
>>>applicable to cuwin mesh?
>>>      
>>>
>>I'm far from an expert in these matters and look forward this weekend to
>>reading some of the material that David referred us to. But I would
>>suggest you not count on any better results in the field than might be
>>suggested by the Kumar article. And it could well be worse. You'll see
>>here some stats on a quick test I helped with in Boston:
>>http://pdos.csail.mit.edu/roofnet/doku.php?id=april2
>>That has encouraged me to think that there and at similar sites, we
>>should aim to use directional antennas on the perimeter as part of
>>trying to minimize hops and perhaps that's something that could be
>>helpful in your situation. Just this past week we've started doing more
>>systematic testing using the Netgear routers so I may have a better
>>understanding within the next few weeks, though they don't lend
>>themselves to swapping antennas (we'll supplement with at least a couple
>>of outdoor boxes that do). Perhaps if you were to describe in more
>>detail the architecture in regard to which nodes would be expected to
>>communicate with which other nodes, David or others on this list could
>>give you a more accurate estimate.
>>
>>Given this description of yours: "here in chicago we're about to deploy
>>a 20 node cuwin mesh backhaul network with cisco APs shooting out to the
>>end users. the ciscos are getting donated fyi but i did get the metrix
>>kits with dual radios for when cuwin can do dual radio"  I think you'd
>>be very intested in the recent segment on Wireless Tech Radio featuring
>>MetroFi founder and CEO Chuck Haas especially from minutes 7:18 to
>>10:00. You can find the entire MetroFi segment highlighted down towards
>>the bottom of the page here:http://www.wirelesstechradio.com/
>>
>>MetroFi use a two radio backhaul system with 45 degree sector antennas
>>and it sounds as if the backhaul radios tend to lock in to a particular
>>path but are aware of each others' location and able to reroute if a
>>backhaul node fails. I'd be interested if anyone knows more about the
>>software that MetroFi uses for that purpose. It seems like the kind of
>>thing that might best suit your purpose at the moment, Stel, if it or
>>something similar is available.
>>
>> Steve Ronan
>>
>>!DSPAM:42d8581b177248678017944!
>>    
>>
>
>  
>