Re: [Roll] reliability metrics for digital radios

[Richard Kelsey <richard.kelsey at ember.com>]

   From: Philip Levis <pal at cs.stanford.edu>
   Date: Mon, 12 Oct 2009 14:54:06 -0700

   On Oct 12, 2009, at 7:26 AM, Richard Kelsey wrote:

   > My point is that the key metric for reliability is not the
   > RTX for a message right now, or any similar metric, but an
   > estimate of the stability of the link itself.  We need to
   > preferentially use stable links.

   If the DAG can reconfigure very quickly, then the need to prefer  
   stable links disappears. E.g., there's a paper appearing in SenSys  
   this year (Bursty Traffic over Bursty Links) that finds using very  
   transient links can reduce path length and improve throughput.

That's an interesting paper, but I think it supports my
point that the core routing needs to use stable links
when possible.  In the paper they set up a base tree
using stable links and then opportunistically skip ahead
using transient links as available.  Distinguishing between
stable and bursty links is the key to what they are doing.

   If the DAG cannot reconfigure or repair quickly, then stable links are  
   important. But by constraining a protocol to only using stable links,  
   we would lose any possible benefits of rapid recovery. This seems like  
   putting the horse before the cart.

I do not think anyone has proposed that we constrain
the protocol to only using stable links.  I do think
that attempting to have at least a tree of stable links
is important.  One possibility would be to have the
preferred parent be one connected by a strong link,
but allow other parents to use less stable, and
hopefully longer, links.  This would allow a middle
ground between reliability and efficiency.

                                -Richard Kelsey