Hi,
Thanks Om.
I was thinking of using receiveEst to centre on a node failure. As you
correctly pointed out, it takes time. Specifically, it took 3 minutes for
the receiveEst to be decreased exponentially. However, other than failure is
there any other case where receiveEst may decrease exponentially, except
say a lossy link.
Secondly, in MultiHopLEPSM, in the updateEst() function, if liveliness
decreases below a threshold, it does: *Nbr->sendEst <<= 1;* However, if it
left shifts sendEst, won't its value increase? Later we check if sendEst <
25, to choose a parent. Isn't this contradictory?
Regards,
Bharat.
On Sat, Jul 25, 2009 at 12:30 AM, Omprakash Gnawali <[email protected]> wrote:
> On Thu, Jul 23, 2009 at 10:52 AM, bharat naik<[email protected]>
> wrote:
> > Hi,
> >
> > Is there any way in which a parent can know if its child has failed?
> > What should the modification to MultiHopLEPSM be?
> > Also, during a simulation, I was failing node 'x' at time 0:5:58.
> > (using Tython). However, even after this, it was being chosen as parent
> by
> > nodes 'y' and 'z'. What can be the reason, since it surely can't send any
> > HELLO packets after being failed? Also, it wasn't as if nodes 'y' and 'z'
> > didn't have any other options.
>
> Those protocols take time to detect that a node has failed. If a node
> does not hear a HELLO mesasge that could be due to node failure or
> lossy link. So a node has to wait long enough. Protocols like CTP rely
> on datapath estimate and can detect these failures quickly.
>
> - om_p
>
_______________________________________________
Tinyos-help mailing list
[email protected]
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
