Early Energy Estimation of Networked Embedded Systems Executing Concurrent Software Components

Abstract

2 Abstract—This paper presents and evaluates a new approach of modeling energy consumption of embedded systems resulted by concurrent software components. The objective is to enable energy estimation within early phases of system development, which allows system designers to compare different allocations of software components within networked systems. The model is presented in detail and its application demo started by a case study. Additionally, an execution time estimation for software components is presented which is necessary for the energy model - but previously not available. The model was developed being applicable early in the development process, i.e. previous to any software implementation. This was realized by using only available information. The individual elements of the presented model are: energy consumption of software components themselves, energy consumption resulted by any software component, and energy consumption resulted by specific software components. The variables of the model can be estimated during early phases of system design using existing methods, expecting the execution time of software components. For that reason, a previously energy estimation technique (1) was further developed to estimate the execution time based on program flowcharts. The estimation was verified by using three commercially available benchmarks. The flowcharts of these are utilized to estimate the execution times. The comparison between estimated and measured execution time of an exemplary embedded system results in an estimation error bandwidth between -12.5 % and +6.8 %. Additionally, an algorithm is presented which enables an automated analysis of program flowcharts as part of the execution time estimation. The developed model was applied within an automotive case study which shows a theoretical energy saving potential of 36.2 %. This demonstrates the potential and relevance of modeling energy estimation within early development phases.