Presentation
The main idea behind the green computing paradigm is reducing the energy cost of executing Sw while keeping the same functionalities and performance. The aim is to lower the electricity demand of any device with computing capacities, or alternatively to enlarge the autonomy of battery-based ones. Despite its importance, Sw sustainability is still in its early stages because of its difficulty, requiring experts with deep knowledge on the underlying Hw. The transition towards green Sw is a must, but it is a highly challenging task that certainly needs research in this direction.
This project proposes the characterization of Sw consumption by means of Multifractal (MF) analysis techniques, generating a signature of the Sw/Hw consumption profile regardless of the execution time, the consumption levels, and uncertainty. The consumption is measured as a high-frequency sampling of the instant input current of the system, building this way a signal (or time series) that we analyze. The final goal of eFracWare is to advance in the characterization of the performance of the binomial Sw/Hw towards a future categorization and certification of the energy efficiency of Sw sustainability, with a solid mathematical background. In order to achieve that, we propose the design and implementation of a novel methodology based on multifractal analysis for the automatic characterization of Sw consumption profiles. It does not directly rely on its consumption values but on its consumption behavior and fluctuations, hopefully providing a metric that is isolated from the underlying Hw. Uncertainty in consumption measurements will be eluded through the analysis of a large number of signals to provide confident results which implies automating the analysis and modifying the methodology typically used in the MF analysis literature. To the best of our knowledge this automatic procedure for MF analysis does not exist yet, and it will be extremely useful for the specialized research community because it can be generalized. Two of the most relevant state-of-the-art MF algorithms will be studied, namely Structure Function Approach (MFSF) and Multifractal Detrended Fluctuation Analysis (MFDFA), based on energy consumption values and fluctuations in consumption, respectively.
This project proposes the characterization of Sw consumption by means of Multifractal (MF) analysis techniques, generating a signature of the Sw/Hw consumption profile regardless of the execution time, the consumption levels, and uncertainty. The consumption is measured as a high-frequency sampling of the instant input current of the system, building this way a signal (or time series) that we analyze. The final goal of eFracWare is to advance in the characterization of the performance of the binomial Sw/Hw towards a future categorization and certification of the energy efficiency of Sw sustainability, with a solid mathematical background. In order to achieve that, we propose the design and implementation of a novel methodology based on multifractal analysis for the automatic characterization of Sw consumption profiles. It does not directly rely on its consumption values but on its consumption behavior and fluctuations, hopefully providing a metric that is isolated from the underlying Hw. Uncertainty in consumption measurements will be eluded through the analysis of a large number of signals to provide confident results which implies automating the analysis and modifying the methodology typically used in the MF analysis literature. To the best of our knowledge this automatic procedure for MF analysis does not exist yet, and it will be extremely useful for the specialized research community because it can be generalized. Two of the most relevant state-of-the-art MF algorithms will be studied, namely Structure Function Approach (MFSF) and Multifractal Detrended Fluctuation Analysis (MFDFA), based on energy consumption values and fluctuations in consumption, respectively.
In order to be able to carry on these research activities a multidisciplinary team have been formed. Some of them are part of GOAL research group at Universidad de Cádiz.
Acknowledgements: This project [TED2021-131880B-I00] has been supported by Spanish national call Transición Ecológica y Digital funded by MCIN/ AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR”.
