cost accounting

AI infrastructure confronts multidimensional physical and economic constraints—including power, thermal management, water usage, interconnect bandwidth, memory capacity, and data throughput—while existing metrics (e.g., PUE, TCO) are siloed and fail to capture the coupled trade-offs among energy efficiency, performance, and cost, hindering cross-layer co-optimization. To address this, we propose a unified measurement architecture grounded in a 6×3 cross-layer taxonomy—spanning facility, network, compute, storage, software, and application layers, each annotated with physical, computational, and economic semantics—and introduce the Measurement Propagation Graph (MPG) to enable, for the first time, system-level, three-dimensional relational modeling. Leveraging systematic literature review, meta-analysis, and graph-based modeling, our framework integrates heterogeneous, multi-source metrics. It supports benchmarking, capacity planning, and total cost of ownership analysis, substantially enhancing interpretability of AI cluster efficiency frontiers and enabling rigorous multi-objective optimization.

Current environmental impact assessments of the AI lifecycle suffer from tool heterogeneity, insufficient coverage of water usage and value-chain stages, and poor cross-study comparability. To address these gaps, this paper proposes a unified operational definition of Green AI and introduces a five-stage lifecycle framework—encompassing hardware manufacturing, model development, training, deployment, and reuse. It explicitly distinguishes Green AI (energy-efficiency–focused) from Sustainable AI (holistically sustainable across environmental, social, and economic dimensions). Methodologically, the framework integrates Life Cycle Assessment (LCA), PDCA-based governance, edge–cloud co-optimized hardware strategies, and a calibrated multi-level metrics system combining estimation and empirical measurement. The resulting methodology enables vendor-agnostic, reproducible quantification of energy consumption, carbon emissions, water use, and embodied impacts. This significantly enhances transparency and comparability, providing researchers, engineers, and policymakers with an evidence-driven, actionable guide for sustainable AI development and deployment.

Current HPC billing mechanisms neglect energy consumption and carbon emissions, undermining user incentives for energy efficiency. To address this, we propose a dual-track transparent pricing mechanism—integrating both energy and carbon footprint metrics—and introduce the first end-user-oriented, multi-resource carbon-aware pricing paradigm, dynamically linking computational cost to real-time energy use and carbon emissions. Methodologically, we combine large-scale power consumption modeling and simulation, a prototype system integrating Slurm with real-time smart meters, and controlled user-behavior experiments complemented by surveys. Our empirical investigation uncovers root causes of weak energy-saving awareness among users and enables the design of an incentive-compatible sustainable computing economic model. Results demonstrate that the new mechanism reduces peak energy consumption by 12–19%, achieves billing accuracy with <3% error, and significantly increases users’ willingness to adopt energy-saving behaviors (p < 0.01).

This study addresses the scarcity of practical approaches for effectively measuring and reducing carbon emissions in real-world computing systems. Conducted on a public cloud platform, this work implements green software practices on an online production service by innovatively integrating the cost-driven mechanisms of serverless architectures with carbon emission estimation. Through energy-efficiency optimization techniques, the system’s carbon footprint is significantly reduced. The project demonstrates that cost optimization can serve as an effective indirect driver for improving energy efficiency. Furthermore, it distills a set of reusable and actionable principles for green software engineering, offering empirical insights and methodological support for sustainable software development in industrial settings.

This study addresses the common oversight of full lifecycle carbon emissions in hardware upgrade decisions by proposing a lifecycle-aware simulation framework. The framework uniquely integrates workload characteristics, location-specific time-varying grid carbon intensity, and multiple embodied carbon allocation strategies—such as uniform amortization and front-loading—with multi-generation CPU power models to dynamically evaluate the total carbon footprint of different deployment scenarios. Experimental results demonstrate that, particularly under low-utilization conditions or in regions with cleaner electricity grids, extending the operational lifespan of existing hardware can substantially reduce overall emissions. These findings challenge the prevailing assumption that newer hardware is inherently more environmentally sustainable and offer a novel paradigm for greener computing practices.

To address the high manual monitoring overhead, coarse-grained modeling, and insufficient integration of carbon intensity (CI) data in scientific workflow carbon footprint assessment, this paper proposes the first automated carbon estimation framework for Nextflow. The framework leverages native execution traces to eliminate manual power instrumentation, dynamically converts energy consumption to carbon emissions by fusing high- and low-resolution temporal CI data, and supports user-defined, CPU-frequency-aware fine-grained power models. Through resource-aware modeling and comparative evaluation against RAPL and GA methods, it achieves precise task-level decomposition of carbon emissions and energy consumption on two real-world Nextflow workflows. Experimental results demonstrate controlled estimation error and superior accuracy over state-of-the-art RAPL and GA approaches. Additionally, the framework enables sensitivity analysis with respect to CI granularity and CPU frequency parameters. The implementation is publicly available.

The ICT sector accounts for 2% of global carbon emissions, necessitating carbon-efficient software engineering; however, existing research is fragmented and lacks systematic integration. Method: We conduct the first systematic literature review (SLR) specifically targeting carbon-efficient software, analyzing 65 state-of-the-art studies through a 5W1H–based taxonomy. Contribution/Results: We propose a unified knowledge structure that rigorously defines the domain’s boundaries, identifies critical research gaps and practical challenges, and synthesizes reusable design guidelines, a standardized carbon footprint measurement framework, and a comprehensive taxonomy of carbon-reduction techniques. Furthermore, we articulate 12 open challenges. This work establishes the first theoretically grounded taxonomy and actionable roadmap for green software engineering, thereby bridging a critical gap in the integration of carbon-aware software development knowledge.

This study addresses the limited scope of traditional high-performance computing (HPC) evaluations, which typically focus solely on performance and energy consumption while overlooking the comprehensive environmental costs of operational configurations. The authors propose the first job-level unified accounting framework that integrates both operational and full life-cycle (embodied) carbon and water footprints. Leveraging life-cycle assessment methodologies, real-time runtime monitoring, and hardware manufacturing emission data, the framework enables fine-grained quantification of environmental impacts. The analysis reveals that increasing thread count generally reduces total environmental footprints, albeit with diminishing marginal returns; while carbon footprints are predominantly driven by operational phases, water footprints are largely dominated by embodied impacts. By jointly incorporating both footprint types at the job granularity, this work establishes a novel paradigm for assessing HPC sustainability.

This study addresses the challenge of optimizing server energy efficiency in high-throughput computing environments, where performance and energy consumption are often at odds. Leveraging real-world operational data and targeted experiments, the work systematically investigates how server configurations influence power consumption, performance, and carbon emissions, uncovering key barriers to implementing effective energy-saving measures in practice. Through empirical power monitoring, workload modeling, and carbon footprint assessment, the authors identify critical factors governing energy efficiency and propose a practical configuration strategy that simultaneously ensures performance guarantees and advances low-carbon objectives. Evaluated under representative high-throughput workloads, the proposed approach achieves substantial reductions in both energy use and carbon emissions.

This study addresses the challenge of inaccurate energy consumption estimation for distributed batch-processing applications like Apache Spark in cloud environments, where node-level hardware energy counters are typically inaccessible. Focusing on Apache Spark deployed on Kubernetes, the work presents the first systematic comparison between resource-utilization-based energy models and ground-truth measurements from Intel RAPL across both AWS bare-metal instances and on-premises clusters. It investigates the impact of CPU and memory utilization signals on estimation accuracy and introduces external monitoring to enhance model fidelity. Experimental results demonstrate that incorporating external monitoring significantly mitigates energy underestimation—reducing the error from −29.58% to −24.41% on AWS and from −24.00% to −16.22% in the local cluster—thereby validating its effectiveness in improving energy estimation accuracy.

This study addresses how interconnection queue constraints in traditional power grids limit AI data center siting, while existing cost assessments often overlook real-time electricity pricing, backup charges, and actual load profiles, thereby failing to inform investment decisions effectively. For the first time, this work endogenizes grid electricity prices and integrates empirical GPU load data, fuel costs, full lifecycle carbon emissions, and U.S. 45V/48E tax credits into a comprehensive site-level cost framework. Employing lifecycle cost analysis, carbon footprint tracking, and investment reverse-engineering, the paper evaluates the economic and environmental performance of nine on-site power generation technologies. Findings reveal that natural gas combined-cycle generation costs ($114/MWh) exceed grid prices ($92/MWh); green hydrogen-based power costs 1.9–2.7 times more than grid supply yet reduces emissions by 85%. On-site generation is fundamentally a product of interconnection capacity and load depth, making cost parity elusive and suggesting current investment strategies may be misaligned.

This work addresses the challenge of accurately quantifying the carbon footprint of scientific workflows in shared virtualized environments, where existing tools rely on oversimplified power models and lack precision. We propose the first high-fidelity carbon footprint estimation framework that supports multi-cluster deployments and is extensible across diverse workflow systems, including Nextflow and Apache Airflow. Our approach integrates workflow execution traces, node-level fitted power models, hardware-level energy measurements via Intel RAPL, and time-aligned grid carbon intensity data, while accounting for operational emissions, embodied carbon, and water–land resource consumption. Experimental evaluation across three clusters demonstrates an average energy estimation error of only 10.8%, substantially outperforming current tools such as nf-core co2footprint, and confirms successful cross-platform deployment.