salle A. Turing CE4
16 January 2014 - 14h00
Estimation of energy consumption of embedded system network
by Ozgun Pinarer from Verimag
Abstract: Embedded system networks are built up by sets of nodes which are communicating with each other. Each of these nodes has some hardware restrictions like limited memory, microprocessor, battery and autonomy. With regard to autonomy, it is crucial to be able to accurately estimate the energy consumption of these systems, both before deployment (for a prediction of life time), and while running (to adapt their behavior depending on the energy level).
The energy consumption of an embedded system depends on how the hardware components are solicited by the software layer. An energy model is conventionally used to characterize each hardware component as a finite set of states. Each of these states represents a given operating mode, which is associated with an instantaneous energy power. To determine the consumption profile of a given node in the network, at any time, the state of consumption of each hardware component should be known. This information can be generated by emulators, capable of delivering highly accurate simulation results (bit-accurate).
In this work, the goal is to propose a solution for estimating and analyzing energy consumption close enough to the actual consumption to derive reliable predictions on the life of the system and to propose an approach that can be implemented on the embedded system in order to provide consumption information at runtime.