EAGER: Cloud Infrastructures for Designing Sustainable Electronics

Information and communication technologies have evolved tremendously over the past two decades to realize modern day internet services, cloud-scale computing, and consumer devices at the edge. While computing forms the backbone of our daily interactions with technology, it also levies a high environmental footprint. Estimates show computing accounts for 4% of the global carbon footprint, on par with the aviation industry’s impact. Furthermore, these emissions are expected to rise given the growing demand for computing devices and scale of services used on a daily basis. Despite the growing scale and importance of computing’s environmental impact, there is a lack of commensurate tools to quantify the footprint. Such tools are prerequisite to effective sustainability-aware optimization. This project aims to leverage cloud resources to build shared community infrastructure and tools to quantify the carbon footprint of computing platforms across their end-to-end lifetimes including hardware manufacturing and operational use. The tools will enable computer systems designers to consider the sustainability implications of systems alongside metrics such as performance, efficiency, power, and energy. The tools are meant to be used in research, sustainability-aware early hardware design, and as pedagogical resources in the classroom. The project will incorporate the tools developed into college courses, as well as in-person and online tutorials and workshops at academic conferences. The tools developed will be open-source, including source code and datasets, to enable further research and development towards a sustainable computing ecosystem. The project aims to develop shared community infrastructure to quantify the end-to-end environmental impact of computing systems. The project focuses on developing methods to consider factors like power and energy, hardware utilization, and the carbon footprint from manufacturing integrated circuits. The project has three major tasks. First, developing carbon models for modern and emerging systems by integrating detailed device characterization, architectural simulators, and product environmental reports of real systems. The models will be parameterized to facilitate researchers to easily add new device characterization of emerging technologies and architectural simulators in the future. Second, building simulators for renewable energy driven systems. The task aggregates carbon intensity datasets of power grids and standardizes output traces from the architectural simulators to enable researchers to consider a breadth of deployment scenarios based on hardware design, and temporal and spatial variability of renewable energy availability. Finally, the third task develops methods to attribute operational and embodied carbon of hardware systems to individual software applications to open opportunities for sustainable application design. To demonstrate the efficacy of the infrastructure, the project includes open-sourcing all software artifacts as well as task-specific case studies, and organizing tutorial and workshops at academic conferences. This project is supported by the National Discovery Cloud for Climate (NDC-C). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.