Moreno Molina, J. (2015). Energy profiling of networked embedded systems [Dissertation, Technische Universität Wien]. reposiTUm. https://doi.org/10.34726/hss.2015.28703
Abstract In this work, strategies to simulate and account energy and power consumption in networked embedded systems are studied, proposed and discussed, in order to enable energy consumption optimization of the system across all levels. All design levels have their contribution and responsibility in overall system energy consumption. However, although energy consumption optimization is a cross-level problem, energy awareness is almost exclusive of hardware design. This is mainly because energy consumption estimation requires accurate power models of hardware subsystems. To bring up energy awareness in other design levels, such as software or network, a holistic cross-level and cross-domain modelling approach is required. Furthermore, openness of wireless, distributed and cyber-physical systems relapses into the cross-domain problem, as it requires considering the significant effect of environment interactions and network events in overall energy consumption. These features are nowadays present in the vast majority of embedded systems networks. Hence, this thesis proposes a simulation framework which is capable of integrating models of all aspects of cyber-physical and distributed embedded systems. The framework is based on SystemC/TLM and C++, which is already an industry standard for hardware and software codesign. In addition, the framework provides a TLM based, inter- and intra-node communication model, including a wireless radio propagation model which enables the integration of the networking aspects. Furthermore, the framework integrates simulation of some physical processes through SystemC-AMS. This thesis also proposes an energy profiling approach to close the semantic gap between the energy consumption estimation, performed at hardware level, and software and network designers. Hardware level data is processed and aggregated in high-level energy profiles that can provide meaningful information that enables energy aware design at software and network levels. The framework is evaluated and contrasted with a state-of-the-art simulator. Framework