cross-layer energy measurement

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.

This study addresses the limitations of existing AI energy consumption assessments, which typically focus on single inference or training runs and fail to capture the real-world energy dynamics of goal-oriented agent systems involving multi-step execution, retries, and recovery. To bridge this gap, the authors propose the A-LEMS framework, introducing two novel metrics: Energy per successful Goal (EpG) and Orchestration Overhead Index (OOI). A-LEMS integrates a cross-layer observation pipeline with a time-bounded attribution model to enable end-to-end, reproducible energy evaluation. Experimental results demonstrate that agent workflows incur an average EpG of 888.1 joules—4.33 times higher than that of linear baselines—while achieving OOI values below 1.0 in tool-augmented tasks, confirming EpG’s sensitivity and effectiveness in reflecting the energy impact of orchestration structures.

This study addresses the critical lack of fine-grained, CPU-level energy observability in edge AI devices, which hinders process-level energy attribution and impedes the advancement of low-carbon AI. Through a systematic evaluation of the ASUS Ascent GX10 platform based on the NVIDIA GB10 SoC, the work reveals that the system only supports instantaneous GPU power monitoring and lacks CPU energy counters and standard power management interfaces such as RAPL, thereby failing to replicate the energy tracking capabilities available on x86 platforms. By combining hardware auditing, reverse engineering of ACPI/SPBM, probing of NVML and SCMI protocols, and calibration with external DC power meters, this research uncovers key blind spots in energy observability across mainstream edge AI hardware—further identifying that MediaTek firmware internally computes per-rail energy consumption but does not expose it. The study proposes hardware requirements for energy-attributable AI and advocates integrating energy observability as a core design metric for AI accelerators.

To address the challenges of low accuracy, high overhead, and slow response in online power estimation under enhanced hardware heterogeneity and increased parallelism for embedded systems, this paper proposes a lightweight system-level power modeling and real-time monitoring method based on Performance Monitoring Counters (PMCs). The method constructs a modular, linear-correlation-driven power model that requires no microarchitectural details and supports flexible, rapid reconfiguration across DVFS states. Integrated with the Linux kernel-level framework Runmeter, it enables low-overhead PMC sampling and runtime power estimation. Experimental results demonstrate an average power estimation error of only 7.5%, energy error of 1.3%, and worst-case kernel monitoring overhead below 0.7%. This enables effective closed-loop task scheduling and workload-aware DVFS control.

This study challenges the conventional assumption that device power consumption is predominantly determined by hardware, instead investigating the influence of user behavior on system-level energy usage. Leveraging Intel telemetry data, the research employs exploratory data analysis and linear regression models to compare power consumption patterns across users in different countries, with a focus on the United States and China. The findings reveal a statistically significant association between user behavior and overall power draw, demonstrating that behavioral factors exert a non-negligible impact on energy consumption. This insight offers a novel perspective for green computing initiatives and provides empirical evidence to inform stakeholders such as Intel in refining energy-efficiency strategies and mitigating environmental impact.

This study addresses the lack of a systematic overview of open-source software energy measurement tools, which hinders energy-aware software design and tool selection. From a mining software repositories (MSR) perspective, the authors employ qualitative content analysis to screen and categorize 585 GitHub projects, identifying 24 high-quality open-source energy measurement tools. The work systematically characterizes these tools in terms of architectural design, measurement granularity—spanning from CPU-level to process, container, and AI workload levels—and their capabilities for carbon emission estimation. By elucidating evolutionary trends in tool development, this research provides software architects with a structured foundation and practical guidance for informed tool selection in energy-efficient software engineering.

研究解决了边缘云连续体中代理AI的能效问题，通过引入agentic-eCAL指标来评估多代理工作流的能耗，从而优化代理位置。

This study addresses the challenges of hardware heterogeneity, system complexity, and the absence of unified evaluation benchmarks in edge-to-cloud deployments of agent applications. To this end, it proposes AgenticOps, a framework that enables automated management across the entire agent lifecycle. The framework establishes an end-to-end pipeline integrating distributed deployment with telemetry collection, and introduces an LLM-as-a-Judge mechanism to facilitate semantic-level automated evaluation and reproducible report generation. Experimental results demonstrate that this approach significantly reduces manual operational overhead while providing a standardized experimental paradigm and an efficient evaluation methodology for agent systems.

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 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.

This study addresses the invisibility of energy consumption in software build pipelines and the limitations of existing estimation-based tools that cannot decompose energy usage across individual build phases. To overcome these challenges, this work proposes an open-source command-line tool enabling hardware-level energy measurement with user-defined phase attribution. By monitoring standard output for fine-grained phase-level energy analysis, the approach transcends measurement constraints inherent to cloud environments. Technically, it integrates Intel RAPL, NVML, and Maven plugins with pattern matching to achieve precise energy tracking. Validation on the Gson project successfully generated comprehensive phase-resolved energy profiles for both cold and warm builds. Ultimately, this research provides a reproducible, fine-grained energy efficiency assessment framework to advance green software engineering practices.