Recent proposals for HTTP-based sustainability disclosure focus on what environmental information should be transmitted at the protocol boundary, but leave open how such per-request values can be generated in realistic deployments. We present a model-based approach for estimating resource consumption and CO2e per HTTP request without requiring fine-grained production power telemetry. Endpoints are benchmarked offline under controlled conditions to derive compact energy models that are evaluated online at the HTTP server boundary. We implement this as an nginx extension that loads a JSON model registry and emits per-request metadata for energy, grid intensity, embodied emissions, and total request-level impact, with only low runtime overhead.

Time- and location-shifting of computational workloads is widely proposed to reduce data-centre emissions by exploiting variation in electricity carbon intensity. However, CO2-only optimization can shift burdens to places where impacts are experienced locally, such as water withdrawals in stressed basins, worsened air-pollution exposure, and increased stress on constrained grids. We present Orca, a sustainability-aware workload shifting framework that jointly considers global climate impacts and heterogeneous local criteria, formulating scheduling as a multi-objective optimization problem. A three-region case study shows that CO2-optimal shifting can worsen local outcomes, while Orca produces context-sensitive schedules that better balance global and local sustainability objectives.

Stakeholders lack a standardized, reproducible way to compare the resource efficiency of software products with similar functionality. We present the Resource Efficiency Software Index (RESI), a measurement-based framework that assigns an interpretable efficiency class (A–E) to products within a software category defined by shared core use cases. RESI evaluates operational resource consumption across four lifecycle phases (INSTALL, IDLE, RUN, UNINSTALL) through a five-step pipeline. We instantiate RESI for ten widely used web browsers, execute the pipeline under a controlled setup, and probe platform effects by rerunning it on a consumer laptop.

MLOps remains a major challenge for individuals and small teams that lack the budget, expertise, or infrastructure to run enterprise-grade systems. MLOX (MLOps-in-a-Box) addresses this gap with an easy-to-use, secure, and fully open-source framework that transforms ordinary virtual or on-premise servers into cohesive, observable, and sustainable MLOps environments. With a service-centric, backend-agnostic architecture supporting Docker, Kubernetes, native, and hybrid connectors, it integrates the full AI workflow—tracking, orchestration, serving, monitoring, registry management, and secret handling—on low-cost hardware while emphasizing transparency, energy efficiency, and reproducibility.

Germany’s commitment to climate neutrality by 2045 is driving scrutiny of the rapidly expanding ICT sector. Based on a survey of 141 participants and targeted interviews with professionals from academia, industry, and politics, this paper investigates the current state, challenges, and barriers to broadly implementing Green Coding in the German software development landscape. Our findings reveal persistent ambiguity around standardized definitions, measurement methodologies, and best practices, while barriers such as insufficient training, limited economic incentives, and organizational constraints impede mainstream adoption.

The increasing energy consumption of software and hardware systems has become a significant concern due to its environmental impact and operational costs. While operating systems like Windows and macOS have proprietary, undocumented energy analytics functionality, this feature is still lacking in Linux. In this paper we introduce PowerLetrics, an open-source framework designed to monitor and analyze power consumption metrics on a process level on Linux.

This talk explores best practices and challenges in measuring software energy consumption, highlighting limitations of tools like Intel RAPL and the Gude Power Meter, and stressing the importance of consistent test conditions, CPU settings, and automated validation. It provides practical advice on specialized operating systems, sampling intervals, and network traffic, aiming to improve green coding practices and reduce software’s environmental impact.

The environmental impact of software is gaining increasing attention as the demand for computational resources continues to rise. In order to optimize software resource consumption and reduce carbon emissions, measuring and evaluating software is a first essential step. In this paper we discuss what metrics are important for fact base decision making. We introduce the Green Metrics Tool (GMT), a novel framework for accurately measuring the resource consumption of software. The tool provides a containerized, con- trolled, and reproducible life cycle-based approach, assessing the resource use of software during key phases. Finally, we discuss GMT features like visualization, comparability and rule- and LLM-based optimisations highlighting its potential to guide developers and researchers in reducing the environmental impact of their software.

Conference publication at HotCarbon 2024 - We present a paper to reduce the energy consumption of LLM inference by using specialized open source models selected by a classifier beforehand.

Veröffentlichung in Future Generation Computer Systems - Das Green Software Measurement Model integriert Ergebnisse von 12 Gruppen von Forschern und Praktikern. In unserem Teil beschreiben wir, wie professionelle Energie- und CO2-messungen mit modernen container-nativen Open-Source-Werkzeugen durchgeführt werden können.

Publication in the GI conference proceedings of EnviroInfo 2023 - Software Life Cycle Assessment (SLCA) is gaining attention for its environmental impacts in production, deployment, usage, and disposal. Unlike LCA for physical products, SLCA is still evolving in software.

Concepts Towards a Life Cycle Assessment of Software - We are sketching out how software emissions over it’s whole lifecycle can be attributed with current tools and methodologies including future challenges and issues.