[Battlemesh] Wireless Battle Mesh v5 - Next steps

Pau hakais at gmail.com
Thu Sep 22 17:34:03 UTC 2011


>
> I do not understand. Once routes are in place it routes traffic. And it
> routes traffic based on the bandwidth. And the bandwidth is the limiting
> factor. And not number of routes existing. No?
>

Yes. But when the device is using 100% of CPU cannot route at the same speed
that when CPU is at 10-20%.
Each mesh protocol have an important overhead on CPU load, and fonera's CPU
is a shit...
So if you are runing 5 mesh protocols inside the same device, the bandwidth
and packet losses are very high, and you cannot test persistent protocols
like TCP.

An example: In our community (guifi.net) we have some mesh clouds. The
biggest one has arround 20 nodes. Normaly we are using ALIX (x86 500Mhz),
and we are able to transfer at 20-30 Mbit/s in 1 hop. The CPU overhead is
not more than 10%.
However ALIX are very expensive, and we tried to use Nanostation5 (which is
like a fonera but a little better). then, running three protocols at same
time (BATMAN, OLSR and BMX) we didn't get a bandwidth higher than 6-7Mbit/s
in 1 hop because the CPU is at 100% all time.
Now apply this to a less powerful device (fonera), using 2.4Ghz (more noise)
and making more than 1 hop. Maybe you get 20-30kbit/s? I really don't know,
but with this kind of bandwidth you cannot test VOIP or Streaming or
Whatever in layer7.

Hey! I made some tests the last night of WBMv4, and my experience was that
with this kind network we cannot have conclusions or a real battle, ICMP are
the only packets that you can use. Nowadays that is not a real "example" of
mesh network.

--
./p4u



2011/9/22 Mitar <mitar at tnode.com>

> Hi!
>
> > It depends on what are you understanding for "run". If run for you is to
> be
> > able to make a ping with less than 50% of losses, then yes is works. But
> if
> > you understand run as a network where you can do layer7 tests, then you
> > cannot say they run.
> > There is several overhead and high cpu load using 5 mesh protocols at
> same
> > time, and fonera is not designed for that.
>
> I do not understand. Once routes are in place it routes traffic. And it
> routes traffic based on the bandwidth. And the bandwidth is the limiting
> factor. And not number of routes existing. No?
>
>
> Mitar
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ml.ninux.org/pipermail/battlemesh/attachments/20110922/73665d2e/attachment.htm>


More information about the Battlemesh mailing list