Section: Research Program
Green IT
With the emergence of the Future Internet and the dawn of new IT architecture and computation models such as cloud computing, the usage of data centers (DC) as well as their power consumption increase dramatically [54] . Besides the ecological impact [68] , energy consumption is a predominant criterion for DC providers since it determines the daily cost of their infrastructure. As a consequence, power management becomes one of the main challenges for DC infrastructures and more generally for large-scale distributed systems.
To address this problem, we study two approaches: a workload-driven [50] and power-driven one [86] . As part of the workload-driven solution, we adapt the power consumption of the DC depending on the application workload, and predict this workload to be more reactive. We develop a distributed system from the system to the service-oriented level mainly based on hardware and virtualization capabilities that is managed in a user-transparent fashion. As part of the power-driven approach, we address energy consumption issues through a strong synergy inside the infrastructure software stack and more precisely between applications and resource management systems. This approach is characterized by adapting QoS properties aiming at the best trade-off between cost of energy (typically from the regular electric grid), its availability (for instance, from renewable energy), and service degradation caused, for instance, by application reconfigurations to jobs suspensions.