[Tinyos-devel] default beacon period for FTSP
kusy at stanford.edu
Sun Feb 1 11:25:27 PST 2009
> On Sat, Jan 31, 2009 at 8:34 PM, Philip Levis <pal at cs.stanford.edu> wrote:
>> On Jan 31, 2009, at 8:21 PM, Omprakash Gnawali wrote:
>>> I suggest a default beacon period of 30s for FTSP. 10s seem excessive.
>> Should FTSP use a trickle timer? The consistency condition would be the
>> linear regression coefficients. Or, better yet, temperature...
> Interpolation might have interesting error properties if the minimum
> interval for Trickle is too short. Consistency condition should also
> include neighboring nodes getting out of sync or there has to be some
> kind of PULL bit.
> Even if the FTSP owners want to stick with periodic beaconing, 10s is
> too frequent.
one problem with ftsp is the long time it takes to initialize (basically
beacon period*3*#hops). i guess the 10s period was chosen for the demo
application to converge quickly. in practice, we never used periods
bellow 30s. i've had success with a 1 hour beacon period (achieving 50us
average and 1.3ms maximum error in a 12 hop network). it required to
increase the period progressively (start with 10sec, then 8min, and then
1 hour period).
thus the trickle timer might be a good idea. it would be interesting to
see some experiments.
More information about the Tinyos-devel