[Roll] Duty-Cycling Radios (was: Re: RPL Metric ID)

[Jonathan Hui <jhui at archrock.com>]

On Oct 13, 2009, at 2:12 AM, Anders Brandt wrote:

> One word of warning on the energy discussion:
> It seems like the assumption is that the routing protocol should  
> manage
> the number of transmitted packets in order to ensure that the energy
> consumption is distributed among more nodes.
>
> There is however a caveat here:
> From our experience, a radio uses energy. Period.
> Receiving uses the same as transmitting - the difference is just which
> amplifiers use the energy.
> Thus, the way to use less energy is to use the radio less - i.e. sleep
> some of the time. Which again translates into higher latency to reach
> such nodes. Obviously, there is something to save, since a radio may
> quickly determine if it is the intended recipient of a packet and  
> return
> to sleep again OR receive the entire packet and stay awake to forward
> the packet on to another node in the network.

You're right.  There is a fundamental tradeoff between energy  
consumption vs. expected communication latency and/or throughput when  
using a particular radio technology.

> Thus, what I am trying to say is:
> Unless the lower layers have a smart way of quickly returning to sleep
> (which some technologies have) the routing protocol will keep all  
> nodes
> awake within direct range - and they will all use power for  
> listening =>
> nothing gained from distributing the traffic.
> Can battery-powered 802.15.4. nodes quickly return to sleep if they  
> are
> not the intended recipient?
> If not, the energy metric may be rather useless (?)

Specific to 802.15.4, it is possible to enable/disable the receiver  
with a period on the order of a couple hundred milliseconds and  
achieve an effective duty-cycle of far less than 1%.  The link simply  
appears as if it has lower throughput.  Expected latency will increase  
proportionally, but the particular factors depends on whether or not  
you can coordinate the duty-cycle schedules to match the routes (or  
vice versa).

In any case, 802.15.4e is working on incorporating at least two duty- 
cycling techniques.  The first is a channel-sampling approach that  
utilizes short chirp packets in the general case but incorporates  
scheduling to reduce transmission overhead.  You can find details of  
the approach being used in Chapter 4 of [1].  The second is a time- 
slotted approach where nodes have to synchronize schedules with their  
neighbors before they can communicate with them.  In both cases, the  
net effect is that they make the link appear as if it is up all the  
time, just with lower throughput and higher latency.  The challenge  
with respect to energy is that they shift energy costs in different  
ways.

[1] http://www.eecs.berkeley.edu/Pubs/TechRpts/2008/EECS-2008-116.pdf

--
Jonathan Hui