May it be for ecological/economical reason or for increasing battery life of our computers and phones, saving power in computer science and networks is a general trade.

During the past years, we have experienced a real effort from chip manufacturers to reduce power consumption of every component of any computational system. The result of this work has been a tremendous increase in battery-life on smartphones and laptop computers.

I have been studying power management in Nvidia GPUs starting from September 2010 in order to improve the efficiency of the GPU when using the open source Linux driver called Nouveau. I think Nvidia GPUs are many years ahead in terms of hardware design and I believe that the concepts introduced by Nvidia are applicable to other systems. This reverse engineering work also helps the research community by publicaly documenting what is otherwise a trade secret and by enabling researchers to experiment with new methods to increase efficiency through the testbed that is the open source driver Nouveau. This reverse engineering work has made possible the writing of Power and Performance Analysis of GPU-Accelerated Systems which has been presented at USENIX2012’s HotPower workshop.

However, no matter how efficient each piece of hardware is, the whole system’s global power consumption can be greatly decreased by coordinating and scheduling operations done by the system. This coordination can involve cooperation between different network nodes in the case a distributed system.

Subsequently, coordination/cooperation among the system’s components raises the problem of security (may it be confidentiality, integrity or availability). Mitigating those issues by designing communication protocols and authentication methods that are safe by design is the only way to limit the power cost of security on these systems. On the other hand, collaboration can also improve security by improving authentication such as described in On Optimizing Energy Consumption: An Adaptative Authentication Level in Wireless Sensor Networks.

During my Ph.D., I am mostly working on Wireless Sensor Networks as a case study for distributed systems and improved local power-efficiency. This study involves hardware, network layers and designing a framework for power-aware collaborative applications.

During my first year, I have been researching on a way to localize communications in Wireless Sensor Networks to reduce power consumption and increase the autonomy of a intrusion-detection wireless network. This work has been funded by the French research agency (ANR) under the project DIAFORUS. The result of this work has been presented at ICAIT2012 under the name Overcoming the Deficiencies of Collaborative Detection of Spatially-correlated Events in WSN.

My current work focuses on designing networks layers that would actually allow the protocol written during the DIAFORUS project to be as efficient as possible. This work involves the CC430 chip from Texas Instrument because of its low power consumption, hardware-support for encryption and highly-programmable radio.

My main objective is to study the interactions between the hardware, network layers and applications. The end goal would be the creation of a simple programming model that would allow applications to be power-aware and that would allow the system to save power opportunistically without disrupting applications.

This thesis is to be presented in 2014.