[OpenWireless Tech] Mesh Networking

[Todd Freeman]

At least we agree on the last part, indeed mesh is good for last mile,
but I think this project should aim for more then a sparsely populated
last mile, I have a response that is much longer that addresses this
issue specifically that is sitting in the moderation queue for being too
long, though I am considering breaking it up into smaller chunks to
bypass that.

No I did not consider multiple physical radios, this seems like it would
get expensive pretty fast if using anything one would consider
SMB/commercial hardware.

on a side note:

    Re: [OpenWireless Tech] Hello World
Is being held until the list moderator can review it for approval.
The reason it is being held:
    Message body is too big: 44775 bytes with a limit of 40 KB

Is it just me or does this limit seem kinda dumb ? Are we really worried
about the list being spammed by messages with long message bodies ?

On 11/02/2012 02:47 AM, michi1 at michaelblizek.twilightparadox.com wrote:
> Hi!
>
> On 01:01 Fri 02 Nov     , Todd Freeman wrote:
>> Mesh networking is pointless when working with more then 5 routers.
> It's hop counts, not routers.
>
>> Its
>> a fundamental part of mesh networking that limits its size, all nodes
>> must divide bandwidth almost equally among all nodes in the chain,  so
>> assuming you start out with 150mbit link, which is VERY generous.  2nd
>> hop, total throughput is 75mbit, 3rd hop, total throughput is 37mbit,
>> 4th hop, 19mbit, 5th op, 10mbit. That is 10mbit for ALL clients to
>> share, starting from 150mbit. Because each of the first 4 hops must
>> carry traffic for its own user as well as the traffic for all the users
>> down the line.
> 150mbit is really *very* optimistic. Especially if you take in mind that the
> collisions which occour when not using directional antennas will slow it down
> a lot.
>
> On the other side the bandwidth is not really correct. For 5 hops, the packet
> needs to be sent and reveived 5 times each. Assuming half duplex transmission,
> and assuming every node is active at the same time, you get the bandwidths
> mentioned above (except it is 15mbit for 5 hops). In practice it is rare to
> see all nodes active at the same time. So these numbers are the "guaranteed"
> bandwidth and the traffic limit and not the actual bandwidth.
>
> Also, current networks do not use the available bandwidth effectively. It is
> similar to the effect on "normal" internet connection "everything else slows
> down if there is one big down/upload running".
>
>> Start out with your home internet and 1mbit upload, you
>> may get to 3 hops and have internet still function if you are very lucky.
> The required uplink bandwidth depends on the number of users and their
> behaviour, not the number of hops they are away. 1mbit at 100% load can carry
> ~300GB/month
>
>> TLDR, mesh networking do no scale at all. Its fundamental in their
>> structure, regardless of protocol.
> Mesh networks might be good for the last mile, but they sure need some kind
> of backbone. Think of them of extending the range of wireless devices rather
> than "internet replacements".
>
> 	-Michi