Energy profiling of large scale applications (Energy)
Leaders: Laurent Lefèvre (RESO), Jean-Marc Pierson (IRIT), Jean-Marc Menaud (ASCOLA)
Green IT as recently emerged as a new research domain. Various researches focus on how to reduce the energy consumption of large-scale infrastructures (data centers, grids, clouds) and by this way how to improve the energy efficiency of these systems.
Our current works show that understanding and modelizing energy consumption of large-scale systems infrastructure and applications are complex tasks depending on vari- ous contexts (location, usage...).
This challenge will focus on the energy efficiency aspects of large scale applications deployed on multiple sites. Understanding the impact of applications on energy usage is a mandatory step in order to propose energy efficient alternatives. Profiling solutions will be proposed to link applications behaviour and features with their corresponding energy usage information. This profiling will cover a set of heterogeneous equipments (computing, storage, networking).
This challenge wil focus on two kinds of targeted large scale applications:
- applications deployed on real physical resources (for low level experiments);
- applications deployed on virtualized resources (virtual machines managed through the ENTROPY software framework).
This challenge will propose some software components able to log, store and expose energy usage to administrators and plateforms users. Preliminary experiments have been done through the help of the ALADDIN energy sensing infrastrucutres (150 sensors on Grid’5000 Lyon site, 6 sensors on Toulouse site and 6 sensors on Grenoble site). Larger experiments involving more sites and heterogeneous equipments (computing and network equipments) will be proposed and analyzed.
Organized events :