[Olsr-dev] OLSR on OLPC?

[David Murray]

Hi Scott
> To quote a Nortel engineer: .......

I think that the point that you make about control traffic being 
affected by data traffic leading to route oscillations is important. In 
the current stable version of OLSR I have noticed that large traffic 
loads will increase the metric. I am guessing that the hello packets 
may occasionally be colliding with data frames or alternatively, the 
hellos are not being transmitted because of a large concurrent data 
frame (these are just guesses).

To me the solution is not really an OLSR problem. AKAIK wireless QoS 
(802.11e) is needed to prioritize these frames. I have had discussions 
with some of the MadWiFi guys and they claim that ad-hoc mode and QoS 
is not easy.

With regards to your comments on metrics
> I provided some links to the 802.11s specs above; do you feel the
> metrics specified are useful?

I think that the 802.11s airtime metric is a pretty good measure. As 
previously stated I think the future plan is to just incorporate 
bandwidth and move to ETT. The current challenge that others are 
working on is how to incorporate layer 2 information into OLSR.

In my opinion, the difference between Airtime and ETT is pretty small. 
They both basically incorporate the bandwidth and reliability. The main 
difference is that Airtime additionally includes some modulation 
specific overheads.

Cheers
Dave