Energy Efficiency in Cloud-Based Big Data Processing for Earth Observation: Current Approaches and Future Directions

The growing concern over the energy costs and carbon footprint of big data processing has sparked significant attention and research in recent years. This focus has largely been driven by the computational intensity of large-language models (LLMs), which are predominantly trained and deployed on cloud platforms. In contrast, despite the substantial data volumes and complex computational demands of Earth Observation (EO) data, energy efficiency in this domain remains underexplored. EO is characterized by an increasing complexity and data volume due to the growing number of satellite missions and the need for multi-scales analysis. This paper investigates the state-of-the-art approaches to energy efficiency in big data processing within the EO domain (EOBD), identifying key challenges and proposing strategic improvements. We begin by outlining the current landscape of big data services in EO, including the prevalent use of cloud computing. We then analyze the unique energy efficiency challenges posed by EO data processing on cloud platforms. This is followed by an examination of common strategies employed to enhance energy efficiency in more widely studied big data environments. Through a literature studies of major EO data processing platforms, we identify significant gaps in energy efficiency practices. Our findings indicate that current EOBD systems often prioritize performance over energy efficiency, leading to unnecessary resource consumption and increased carbon emissions. Finally, we propose potential avenues for future research, including the development of benchmarking frameworks and energy measurement toolkits. We also suggest optimization techniques for infrastructure orchestration and task assignment within distributed cloud-based processing frameworks. These proposed approaches aim to foster more sustainable practices in EOBD, potentially reducing energy consumption and environmental impact while maintaining processing capabilities.