[MLB-WIRELESS] meshing

[Andrew Harcourt]

G'day!

> I have this impression, that somewhere in the back of our
> collective, Melbwireless hivemind, we have these vague
> thoughts that at some stage we are going to have a mesh
> network.  By that I mean that nodes will be able to pass
> packets along from one to the other, hippity-hopping their
> way from one side of Melbourne to the other.

I don't have time right now to write a full response to this, but I doubt
this is such a wonderful goal to aspire to.

If person B is smack bang between persons A and C, they're going to do a lot
of packet forwarding. If person M is between persons A-L on one side and N-Z
on the other, they are going to be doing an *extreme* amount of packet
forwarding.

I think that this kind of approach will quickly swamp the geographically
centralised nodes and cause their maintainers to lose interest - hey,
wouldn't you do the same if you couldn't get a single packet of your own
out? Sure, you can use traffic shaping and a whole bunch of QoS rules, but
at the end of the day if you're forwarding heaps of traffic for others and
not getting much out of it yourself, you'll start to lose enthusiasm.

Contrast this with a backbone-style setup. I'm sure there are a few
(dedicated? altruistic? bloody minded?) folk around who are happy to run a
few point-to-point links as a backbone and then hang other branch nodes off
them. The routing issues will have to be sorted out a little more
thoroughly, but it will give us more overall bandwidth to play with if we
use a backbone-style topology.

> I think this is something we need to discuss if it is to be
> a long-range goal for the network.  Maybe we can make
> provisions in advance for this capability at geographically
> significant sites. This would prevent us from investing in
> unsuitable hardware, and cursing that choice down the road.

IMHO we're going to end up with two main types of nodes: 1) leaf nodes and
2) everything else.

Leaf nodes could probably (quite happily) use a normal AP plugged straight
into an Ethernet hub to connect to a branch node with an omni. Leaf nodes
wouldn't need any kind of routing hardware or software - just run them as an
Ethernet bridge and all should be well. This is the simplest way we can get
other people involved - they only have to buy a single piece of hardware, it
comes in a box and they can just plug it in and connect.

Other nodes are going to have to handle a few more routing issues and the
odds are that we'll have to run a decent OS such as Linux or *BSD (no
religious flames please) on the routing nodes.

Out of time - have to go. If anyone else has thoughts on the matter, please
yell out.

Regards,
Andrew


--------------------------------------
Andrew Harcourt
Telstra Research Laboratories

+61392536210 (office)
+61409723311 (mobile)
andrew.harcourt at team.telstra.com
--------------------------------------




--
To unsubscribe, send mail to minordomo at melbwireless.dyndns.org with a subject of 'unsubscribe melbwireless'  
Archive at: http://melbwireless.dyndns.org/cgi-bin/minorweb.pl?A=LIST&L=melbwireless
IRC at: au.austnet.org #melb-wireless