On Apr 13, 2008, at 3:13 PM, John Frye wrote:
> Hello,
>
> I have several motes collecting data and communicating through the  
> provided collection and dissemination protocol implementations in  
> the net library.  However, the battery life on these motes is quite  
> poor; they only last for around a week and half.  I attempted to  
> implement low power listening as shown in the tutorial, but to no  
> significant gain.
>
> Consequently, I was thinking about tweaking the collection  
> implementation to see if there was any sort of possible gain.  How  
> is the routing table updated? Are period messages sent to neighbors?  
> If so, is there a constant or value I can modify to decrease the  
> frequency of these messages?
>
> I would very much appreciate answers to these questions or other  
> suggestions.

The version of CTP in CVS is not well designed for low-power, because  
it uses excessive control traffic. The work up to 2.0.2 (and what's in  
CVS) focused on making the data path as efficient as possible: in many  
cases CTP sends 25% fewer data packets than MultihopLQI, with either  
equivalent or much higher reliability. Not much attention was paid to  
how often control packets were sent.

There is work in net2 going on right now to bring down the control  
traffic: the test version (not checked in) cuts control traffic by a  
factor of 10. When the network starts up, it sends more control  
packets as it discovers the topology and adapts. Over a few hours, it  
settles down. In our experiments, it sends more control traffic than  
MLQI for the first ten minutes or so, but after a few hours it's  
sending at 1/2 -1/4 the rate.

Om is working on this right now, and David Moss has been examining how  
it interacts with LPL. We hope to check in the changes in the next few  
weeks.

Phil
_______________________________________________
Tinyos-help mailing list
[email protected]
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help

Reply via email to