Interest in wireless networking technology has been increasing over the last decade.
To cope with the demand, network design has been focusing on increasing the throughput,
at the cost of increaed energy expenditure. One of themajor challenges for current
wireless networks is offering the required performance while keeping the energy consumption
low by adapting the system configuration to the varying environment and
application constraints. This is a challenging task that requires a global approach considering
both the application and the hardware characteristics. Also, both performance
and energy are affected by the networking protocols and the behavior of other users.
In this doctoral thesis, a framework is derived to jointly optimize energy and performance
for a subset of relevant networking scenarios in which nodes interact. Methods
are given to take into account complex hardware, and how to optimize energy versus
performancewith that hardware under specific application and networking constraints.
Next, the framework is instantiated in two relevant and practical contexts for wireless
networking and contributions are brought to both of these contexts.
First, an 802.11a Wireless Local Area Network is considered. A Methodology for
Energy-Efficient Resource Alocations (MEERA) is derived to optimize the energy for
a centrally controlled network with delay and loss sensitive video traffic. By means
of effective run-time adaptation, MEERA outperforms state-of-the-art energy management.
MEERA is also implemented succesfully to control a real HIPERLAN/2
system. Next, a solution is derived to optimize performance and energy in a network
where nodes locally choose their configurations and channel access. It is shown that
local energy management on top of 802.11 results in a Tragedy of the Commons. We
then propose a fully distributed solution based on Shortest Job First scheduling that
effectively optimizes the performance or energy depending on local preferences.
Second, we instantiate the framework for the very different IEEE 802.15.4 context
of low-rate and low-power sensors. The MEERA approach can easily be adopted
for clustered sensor networks, if the constraints and system models are replaced to
represent the new design goals. For the distributed channel access, we note that the
802.15.4 is very different then 802.11, and we propose a newMarkovmodel that mimics
this behavior. With the model, we show that distributed energy and performance
optimization for 802.15.4 results in an acceptable equilibrium.
Finally, we contributed to the field of cognitive radio to evaluate the framework for
the design of this disruptive technology.