Digital Signal Processing Reference
In-Depth Information
be achieved for a network of users with bursty delay-sensitive data and over a slow
fading channel
.
We briefly recapitulate the following three observations that show that the pro-
posed adaptation problem is not trivial and requires to streamline energy manage-
ment approaches across layers.
•
First, state-of-the-art wireless systems such as 802.11a devices are built to func-
tion at a fixed set of operating points assuming worst-case conditions. Irrespective
of the link utilization, the highest feasible transmission rate is used and the power
amplifier operates at maximum output power [21]. For non-scalable systems, the
highest rate results in the smallest duty cycle and hence the lowest energy con-
sumption. On the other hand, for scalable systems, this strategy results in ex-
cessive energy consumption for average channel conditions and link utilizations.
Recent energy-efficient wireless system designs focus on VLSI implementations
and adaptive physical layer algorithms where a lower transmission rate results in
energy savings per bit [38, 71]. For these schemes to be practical, they should be
aware of the hardware energy characteristics at various operating points.
•
Second, to realize significant energy savings, systems need to shutdown the com-
ponents when inactive. This is achieved only by tightly coupling the MAC com-
munication to the power management strategy in order to communicate traffic
requirements of each user for scheduling shutdown intervals.
•
Finally, intrinsic trade-offs exist between schemes while satisfying the timeliness
requirements across multiple users. As the channel is shared, lowering the rate of
one user reduces the time left for the other delay-sensitive users. This forces them
to increase their rate, at the cost of energy consumption or bit errors.
Our approach couples these tasks in a systematic manner to determine the optimal
system-wide power management at run-time.
6.1.2 Smart Aspects and Energy Efficiency
We propose a two-phase run-time/design-time solution to efficiently solve the sleep-
scaling trade-off across the physical, link and MAC layers for multiple users. It is
an instantiation of the abstract design flow discussed in Chap. 3. At design-time the
problem is resolved by searching for a set of close-to-optimal points in the solution
space. This anticipates a set of possibly good system configurations. Starting from
this configuration space, we can schedule the nodes at run-time to achieve near-
optimal energy consumption with low overhead. The flow is illustrated in Fig.
6.1
.
It is clear that the set of possible points is fully characterized at design time. At run
time, as function of the monitored scenario, the most energy efficient configuration
can then be selected. In the next section, the design approach is detailed further.
The research in this chapter has been performed in close collaboration with Rahul
Mangharam at Carnegie Mellon University. This cooperation also leads to several
shared papers [72-75].
