[SensorNetArch] MAC + Network Layer Idea

[Philip Levis]

On Nov 5, 2004, at 2:32 PM, David Culler wrote:

> I would love to discuss this paper at one of our meetings.  For me it 
> was the single most important paper at Sensys.

I think it would also be good to look at Byers' paper to understand the 
bitfield result better; he had a paper in ICDE 2004 which is a 
precursor to the results in Suman's work. One of the authors gave a 
talk at ICSI earlier this year, and I was impressed; it was a very 
theory oriented paper, but all of the assumptions it made about what 
motes could do were right on, so it was clear that the approach could 
be easily implemented. Theorists in the real world, etc.

On a tangent... I'm starting to think that whole-network aggregation is 
oversold; at some point, it seems counterintuitive to deploy 8000 nodes 
in order to collect sixteen bits of temperature information. My thought 
is that aggregation is extremely important for *management.* Finding 
misbehaving nodes is a good use of a top-K query (i.e., GDI's 
conclusion that erratic sensor readings are a sign of imminent node 
failure). Synopsis diffusion shows that you can make whole-network 
aggregation simple and lightweight, which is one of the main 
requirements for a management system.

Phil

-------

"We shall not cease from exploration
And the end of all our exploring
Will be to arrive where we started
And know the place for the first time."

- T. S. Eliot,  'Little Gidding'