Architecting for Sustainability: A Software Architects Guide to Green Software Keynote by Eoin Woods

ICT accounts for 2-4% of global greenhouse gas emissions, comparable to aviation, and data center electricity use is projected to nearly triple by 2030. Software architects can reduce emissions through demand shifting (moving workloads to times/regions with cleaner grids), demand shaping (varying computation intensity based on grid carbon intensity), offloading work to client devices, and measuring emissions using tools like cloud vendor dashboards, Cloud Carbon Footprint, and Electricity Maps. For AI systems, key levers include using smaller pre-trained models, quantization, and scheduling training during low-intensity grid periods. A real-world case study shows a machine learning team on Azure nearly halved emissions by automatically routing workloads across regions based on real-time grid intensity data from Electricity Maps combined with the Green Metrics Tool. Practical starting points: learn the basics, eliminate unnecessary compute and storage, understand your grid carbon intensity, then consider architectural changes before optimizing code.