[MLB-WIRELESS] meshing

Roger Venning r.venning at telstra.com
Mon Jan 21 14:48:34 EST 2002


David & the list,

I would dispute that ad-hoc mesh networks have 'poorly understood'
routing behaviour. The IETF Mobile Ad-hoc Networks working group (MANET
- http://www.ietf.org/html.charters/manet-charter.html) has been going
for many years now, and embodies considerable research in the field. The
particular instance of an ad-hoc network that an operational Melbourne
Wireless would represent is almost a trivial case given the distinct
lack of mobility and the reasonably large link lifetimes that would be
expected. This doesn't mean that the problem is simple though.

One of the largest issues involves not connectivity - the simplest goal
of any routing algorithm that could be run on the network - but rather
traffic engineering, the 'art' of getting load balancing across not just
multiple links but multiple paths. This is a problem that is critically
dependant upon traffic dispersion - where the traffic sources and sinks
are. It will be a big issue if there is for instance one particularly
popular resource within the network - e.g. a Internet connection,
wormhole to another wireless network, etc.

One interesting thing to note about 802.11 is that the medium access
control layer is not optimal for supporting peer-to-peer forwarding as
found in a mesh network situation. For more information have a look at a
paper that was published last year in the IEEE Communications Magazine -
http://www.cs.ucsb.edu/~cs290i_mc/papers/80211_adhoc.pdf.

I've been thinking for a while now about the need to bootstrap ourselves
up from a very low density network towards a network of a 1000-10000
nodes. I think it is essential that the solution that serves the network
to begin with be low cost - just one wireless interface. I am almost
certain though that any architecture that serves the low density
starting point will be adequate for a 'successful' Melbourne Wireless.
However we must bridge that river when we come to it. I think I remember
an anecdote about some IT supplier selling itself on the basis of how
Amazon couldn't have scaled to their current size without them - and the
retort that if Amazon tried to start with the complete system solution
from this vendor that they never would have succeeded.

I'll 'publish' my thoughts & software kit on how to drive either an
access point or a wireless interface from a PC class router node in a
way that provides service to local wireless clients as well as
supporting connectivity to the mesh Real Soon Now (TM). The intent is
that this will be a model that can get us started with a low density
network. (Now I've put that on the table I'm guessing I'll be trying
harder to deliver the goods :-) ).

Roger.

----- Original Message -----
From: David Arnold <arnold at dstc.monash.edu.au>
Date: Monday, January 21, 2002 2:02 pm
Subject: Re: [MLB-WIRELESS] meshing 

> -->"Jason" == Jason Hecker <jason at air.net.au> writes:
> 
>  Jason> Also, the more antennas that go up, the more flexible the
>  Jason> routes as well.  So if someone goes down, the mesh can
>  Jason> automatically readapt and deal with it.
> 
> let's distinguish between an ad-hoc mesh, where all nodes are routers
> and use omni-directional radio, and a configured backbone architecture
> with redundant uni-directional links.
> 
> such a backbone architecture gives you nearly as much resilence, with
> a much better chance of getting reasonable bandwidth, fewer problems
> of people having their bandwidth sucked unwillingly, and will work
> with well-understood routing algorithms.
> 
> pure ad-hoc mesh networks have poorly understood routing behaviour,
> less predictable (and manageable) bandwidth usage, and require a radio
> setup that is more prone to interference as well.
> 
> 
> 
> 
> 
> d
> 
> --
> 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=melbwirelessIRC at: au.austnet.org #melb-
> wireless
> 
 


--
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



More information about the Melbwireless mailing list