[Battlemesh] Meshing with strong/near clients and weak/far nodes

[Saverio Proto]

Hello Ben,

it is a well known problem of WiFi. 1 weak client breaks the all
thing. It is known as 802.11 Performance Anomaly.

https://www.it.uu.se/edu/course/homepage/datakom2/vt07/Seminars/802.11Anomaly_Infocom03.pdf

this paper is from 10 years ago, but is still a good paper.

Saverio

2014-02-18 18:31 GMT+01:00 Ben West <ben at gowasabi.net>:
> Thank you for the pointer to JaldiMAC!  Looks it is/was indeed a rough
> equivalent of AirMAX for 802.11n Atheros radios.
>
> Unfortunately, I think the JaldiMAC trail goes cold here.  No updates to
> this repo I found since 3 years:
> https://github.com/shaddi/jaldimac/commits/master
>
>
> On Tue, Feb 18, 2014 at 11:10 AM, fboehm <fboehm at aon.at> wrote:
>>
>> Am 18.02.14 17:40, schrieb Mitar:
>>
>>> Hi!
>>>
>>> I think using TDMA would help here. Each client would reserve its time
>>> slot and it would not matter how strong they transmit in their time slot.
>>>
>>> Sadly open source drivers do not (yet) support TDMA. But Ubiquiti ones
>>> do.
>>>
>>>
>>> Mitar
>>>
>>>> I'm curious if list members have found clever approaches for operating
>>>> mesh
>>>> nodes that must talk simultaneously to near/strong clients and far/weak
>>>> clients.  I've found that, at least with atheros / ath5k-based nodes
>>>> running OpenWRT AA r39154, which do single-radio meshing via multiple
>>>> VIFs,
>>>> connecting a particular strong client to the AP VIF of one node can
>>>> disrupt
>>>> that node's communication with a distant node on the adhoc VIF.
>>>> Admitted,
>>>> the ath5k driver is pretty suboptimal, but its weakness does at least
>>>> demonstrate the nature of the problem easily.
>>>>
>>>> This problem is less pronounced with ar71xx-based nodes, e.g. UBNT
>>>> Airmax
>>>> gear, and I do understand it is going to fundamental to single-radio
>>>> meshing applications anyway (i.e. hidden node problem).  Likewise, one
>>>> can't always control clients' TX strength, besides mounting nodes at
>>>> locations such that a client can never be physically too close, e.g. on
>>>> a
>>>> rooftop or high on an exterior wall.
>>>>
>>>> Acknowledging the limits of the radios, I'm curious if anyone has had
>>>> reasonable success in these situations with things the following:
>>>>
>>>> - RF-Armor style shielding on the back of the mesh node, i.e. to reduce
>>>> the
>>>> influence of stray noise or reflections
>>>> - Setting an RTS value
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Battlemesh mailing list
>>>> Battlemesh at ml.ninux.org
>>>> http://ml.ninux.org/mailman/listinfo/battlemesh
>>>>
>>>
>> It's quite a while since I read it but you might want to have a look at
>> "JaldiMAC". It's kind of a 802.11 based polling protocol.
>>
>> http://matthias.vallentin.net/papers/nsdr10.pdf
>>
>> By the way. As far as I know AirMAX protocol from Ubiquiti Networks is
>> also based on polling the clients. Although it seems to be a quite advanced
>> and mature polling protocol.
>>
>> Franz
>>
>> _______________________________________________
>> Battlemesh mailing list
>> Battlemesh at ml.ninux.org
>> http://ml.ninux.org/mailman/listinfo/battlemesh
>
>
>
>
> --
> Ben West
> http://gowasabi.net
> ben at gowasabi.net
> 314-246-9434
>
> _______________________________________________
> Battlemesh mailing list
> Battlemesh at ml.ninux.org
> http://ml.ninux.org/mailman/listinfo/battlemesh
>