Over the last twenty years, energy efficiency has become an essential concern for IT. Not only because high-performance processors are constraint by energy delivery and heat dissipation, but also owed to an assortment of factors that transcend the scope of IT itself. Climate change is the most obvious, since operating large data centers (DC) requires a massive amount of electricity whose generation contends with the goals of the Paris Agreement signed in 2015. After generation, electricity delivery becomes the next issue straining electric grids worldwide; yet electricity supply is also important for IoT and wearable devices, which are frequently operated on batteries of limited capacity and lifetime. Energy efficiency also limits computational density in high-performance environments, since computing devices convert virtually 100 % of the electricity delivered to them into heat—which needs to be extracted first from the devices themselves and then from the DC. On the other extreme of IT, for the billions of IoT sensors, wearable andpersonal devices distributed around Earth, autonomy is a critical aspect hanging on batteries, accumulators and energy harvesters. Finally, we should not forget that computing inefficiencies carry the need of additional devices, which exacerbates energy demands across the lifecycle of those additional devices, i.e., for raw material extraction, processing, manufacturing, transportation, recycling and disposal. In this talk, we will assess the implications of IT on the global energy consumption landscape and we will explore avenues to maximize the energy efficiency of IT operations. As DC operators, should we renew computing resources or extend their lifetime? The answer depends on the carbon intensity of the electricity grid supplying the DC, the embodied emissions of the equipment, and the jump in performance expected from the new hardware. Should we instead plunge into a reduction of the carbon intensity of the grid? Or should we invest our efforts on the design of more efficient hardware architectures? We will explore these and other questions from the perspective of EcoCloud, a center of EPFL for research on sustainability.

TBD

The goal of the E2CC project is to develop a standardized, end-to-end platform for supporting edge-cloud applications. It aims to guarantee interoperability, portability, infrastructure management, and orchestration while integrating decarbonization, AI/MLOps, and cybersecurity. E2CC is one of 19 IPCEI-CIS projects (Important Project of Common European Interest on Next-Generation Cloud Infrastructures and Services). It contributes to several open-source projects, including ALUMET and MojitO/S, which monitor energy consumption to help optimizing decarbonation.

The time, complexity and data intensity associated with the Life Cycle Assessment (LCA) of electronic systems makes it impractical to use during design and support design choices. This talk will present STEETCH, an eco-design tool made by and for electronic designers, leveraging insights from interviews with electronic experts. STEETCH consists of an open-access web application relying on simplified LCA, design guidelines and educational content to facilitate the introduction of environmental metrics from the early stages of design in order to support decision-making.

TBD