Accountable carbon footprints and energy profiling for serverless functions
Reading group: Artur Pereira Galvão will present "Accountable carbon footprints and energy profiling for serverless functions" (SoCC'24) at 1D23 the 11/12/2026 at 11h30.
Abstract
Cloud computing is a significant and growing cause of car- bon emissions. Understanding the energy consumption and carbon footprints of cloud applications is a fundamental prerequisite to raising awareness, designing sustainability metrics, and creating targeted system optimizations. In this paper, we address the challenges of providing accurate and full-system (not just CPU) carbon footprints for serverless (FaaS) functions. To the best of our knowledge, this is the first work which develops an energy and carbon metrology framework for FaaS. Carbon footprints require a new approach to energy profil- ing. We use FaaS workload properties such as locality to de- velop a simple and practical online statistical disaggregation approach. Our fine-grained per-invocation carbon footprints also include shared hardware and software emissions, and use insights from Shapley values to fairly account for both operational and embodied emissions. Owing to the growing importance of carbon measurement, we develop a new rig- orous marginal energy based validation methodology which results in accountable, complete, and fair footprints. Over a wide range of FaaS workloads and hardware platforms, our energy footprints have an accuracy of > 99%.
