---
title: "METRION: A Framework for Accurate Software Energy Measurement"
description: "Information and communication technology accounted for about 1.4% of global greenhouse gas emissions and 4% of the world's electricity use in 2020, and both…"
url: https://greencode.ai/publications/metrion-a-framework-for-accurate-software-energy-measurement/
---

# METRION: A Framework for Accurate Software Energy Measurement

- **Authors:** [Benjamin Weigell](https://orcid.org/0009-0006-8362-2746 "ORCID profile of Benjamin Weigell") , [Simon Hornung](https://orcid.org/0009-0005-9033-3013 "ORCID profile of Simon Hornung") , [Bernhard Bauer](https://orcid.org/0000-0002-7931-1105 "ORCID profile of Bernhard Bauer")
- **Contributing partner:** [University of Augsburg](https://greencode.ai/partners/university-of-augsburg/ "University of Augsburg in GreenCode")
- **Published in:** Proceedings of the IEEE/ACM 10th International Workshop on Green and Sustainable Software (GREENS '26), Rio de Janeiro, Brazil, ACM, 2026, pp. 131-138
- **Published:** Apr 2026
- **Type:** Conference paper
- **DOI:** [10.1145/3786148.3788635](https://doi.org/10.1145/3786148.3788635)

### Synopsis

Information and communication technology accounted for about 1.4% of global greenhouse gas emissions and 4% of the world's electricity use in 2020, and both are expected to grow. Reducing that through software optimisation needs two things first: knowing which applications use the most energy, and being able to check whether a change actually saved any. This paper introduces an energy attribution model that assigns CPU and DRAM energy to applications at the level of individual threads, allowing for simultaneous multithreading, frequency scaling, multi-socket machines and non-uniform memory access. The model is built into METRION, an extensible framework with a platform-independent data model and a first implementation for Linux on Intel CPUs. Tested on three workloads, it captured the CPU energy of CPU-bound applications with a mean absolute percentage error of 4.2%, and the DRAM energy of memory-bound applications with an error of 16.1%. Accurate per-application measurement of this kind is what makes energy savings in software verifiable rather than assumed.

 [Download PDF](https://doi.org/10.1145/3786148.3788635) [View at the publisher](https://doi.org/10.1145/3786148.3788635) [All publications](https://greencode.ai/publications/)Cite this work

Weigell, B., Hornung, S., & Bauer, B. (2026). METRION: A Framework for Accurate Software Energy Measurement. Proceedings of the IEEE/ACM 10th International Workshop on Green and Sustainable Software, 131–138. <https://doi.org/10.1145/3786148.3788635>
