[OLPC Networking] Application awareness in the mesh network?

[Krishna Sankar (ksankar)]

Yep, we would need the topology awareness as well as power aware
routing, not only in the end point laptops but also all the laptops who
are in the path in the mesh.
 
>	Then again, I might be thinking too much into a problem that
doesn't exist.
<KS>
	Kind of. Usually the mesh routing finds the shortest path than
the applications. The reason is this can change depending on the devices
who come in and go out of the network, interferences et al. While
topology aware applications are interesting, very difficult practically.
Having said that, if the apps need this information, we can percolate
the topology proximity and other metrics, thru some appropriate APIs.
</KS>
 
Keep asking questions, we need to document them so that as we progress
we can start implementing them ...

Cheers
<k/>

________________________________

	From: networking-bounces at laptop.org
[mailto:networking-bounces at laptop.org] On Behalf Of MBurns
	Sent: Wednesday, September 06, 2006 3:08 PM
	To: networking at laptop.org
	Subject: [OLPC Networking] Application awareness in the mesh
network?
	
	
	Since I am in the asking questions mood... 
	
	The OLPC will (as mentioned in the previous thread) have a mesh
network at L1-3. And given the nature of these devices, many of the
applications (roughly L4-7) will also be distributed by their nature.
IM, chat and collaboration will have to be p2p, due to the questionable
Internet connection/bandwidth. 
	
	Essentially, that will have a p2p system (with geographical
awareness due to wireless signal, at L1-3) creating a 'flat' IPv4/6
network. On top of that we will have p2p systems (wiki, mDNS, IM) that
form their own connections and structure. 
	
	With two p2p systems, one physically con trained and one
logically constrained, sitting on top of the other, it is possible to
have a very inefficient path to information, putting unneeded strain on
the network, and leading to degraded service, at worst. To some extent,
this is unavoidable (IM'ing to your friend who happens to be on the far
end of the wireless network), however, is it reasonable/possible to try
and put topology awareness in the applications to encourage them to
structure in such a way as to maximize efficiency in terms of the
wireless mesh? 
	
	Then again, I might be thinking too much into a problem that
doesn't exist.
	
	Michael Burns
	Network Engineering
	Oregon State University