On trade-off in software engineering: a way toward sustainability

June Sallou has a PhD in Software Engineering from the University of Rennes 1, France. She also has a background in Agronomy and Bioinformatics. Her research interests include Scientific Modelling, Approximate Computing, Environmental Science, and ICT for sustainability.

Abstract

At a time when climate change is a major issue, there is a need to put sustainability at the centre of the discussion in order to reduce the increasing overconsumption of resources and its associated consequences. As “software eats the world”, software engineering has a role to play in this task, both in more sustainable software practices and in the use of software to enable more sustainable societal practices. Thus, adapting software practices to ensure relevant and reasoned consumption of computational resources according to the purpose of the software and its context of use seems appealing. Trading-off some accuracy for better performance (e.g., execution time, energy consumption) by reducing the number of computations is one approach to achieve this. To demonstrate the interest and potential of such an approach, I will present two application cases. The first deals with the application of approximate computing on a scientific simulation model in order to enable real-time and more sustainable decision-making by policymakers in the area of climate change. The second presents the potential of applying data-centric approaches in the case of Green AI in order to reduce the extensive amount of energy consumed when training algorithms used in the field of artificial intelligence. In both cases, promising results have been obtained, encouraging further research on software techniques to automate, generalize and validate the practice of trade-offs in software implementation and execution, and also to extend and create new opportunities for software to be used with regard to sustainability.