Process-level Power Estimation in Multi-core Architectures

Maxime Colmant (1, 2, 3) , Romain Rouvoy (3, 1) , Lionel Seinturier (3, 1)

1 SPIRALS - Self-adaptation for distributed services and large software systems
2 ADEME - Agence de l'Environnement et de la Maîtrise de l'Energie
3 Université de Lille, Sciences et Technologies

The ICT has an huge impact on the world CO2 emissions and recent study estimates its account to 2% of these emissions. This growing account emissions makes IT energy efficiency an important challenge. In this article, we describe a solution which is able to estimate the software power consumption for multi-core systems. An application-agnostic power model is built automatically following a specific approach. Our solution, PowerAPI, based on actor model, uses this power model to estimate the software power consumption at runtime. With the help of runtime metrics, PowerAPI allows to get accurate power estimations, without requiring an important hardware investment. The experiments shown that PowerAPI is a reliable and a non-invasive solution.

Domaines: Informatique [cs]; Génie logiciel [cs.SE]

Maxime Colmant, Romain Rouvoy, Lionel Seinturier. Process-level Power Estimation in Multi-core Architectures. Compas, Jun 2015, Lille, France. ⟨hal-01171704⟩