Optimizing the energy efficiency of information systems: a comparative study of data delivery architectures from Agrosyst

Abstract

The increase in energy consumption by data centers, which is projected to double between 2024 and 2026, and more generally the growing impact of digital technology on the climate, justify the need to consider the best technological solutions regarding their energy consumption. As part of the evolution of the methods for providing agricultural data from more than 2,000 farms and experiments in the DEPHY network, stored in the Agrosyst Information System, to researchers and data scientists, the issue of optimizing energy efficiency has emerged. Two types of data delivery architecture are being compared: a traditional architecture characterized by the systematic storage of final tables, and an architecture characterized by the creation of numerous intermediate tables and an “on-demand” generation process. The study shows that the choice between these architectures depends on the volume and specificity of the queries. Given the current usage patterns of the Agrosyst system, it seems justified to change the architecture, thereby reducing server storage and favoring the execution of certain procedures at each instance of data generation. This paradigm shift could reduce the energy consumption of the data provisioning system by at least half. We believe that the methodology used for this analysis has a broad applicability and may show potential interest for other similar situations. This work lays the foundation for systematically considering energy efficiency in future developments of the Agrosyst Information System and could support the development of a more robust and generalizable methodological framework.