Where Should Agents Live? Energy-Memory Characterization of Agentic AI for the Edge-Cloud Continuum

📄 arXiv: 2609.18283v1 📥 PDF

作者: Carolina Fortuna, Vid Hanžel, Tim Strnad, Blaž Bertalanič

分类: cs.AI, cs.LG, cs.MA

发布日期: 2026-09-16

💡 一句话要点

提出agentic-eCAL以优化边缘云环境中的智能体AI能耗与内存管理

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 智能体AI 能耗管理 边缘计算 多智能体系统 5G网络 AI生命周期 数据传输 性能评估

📋 核心要点

  1. 现有的AI生命周期指标仅关注单一模型推理，未能有效评估多智能体工作流的能耗，导致网络运营商缺乏决策依据。
  2. 本文提出agentic-eCAL，结合闭式两速单调用能量模型与OSI数据传输，针对多智能体工作流的能量成本进行全面评估。
  3. 实验结果显示，智能体间文本传输的能耗占比极低，主要能耗来自推理和上下文处理，提供了智能体放置的新视角。

📝 摘要（中文）

随着电信网络向自主5G-Advanced和6G操作演进，智能体人工智能（AI）工作流在边缘-云连续体中日益普及。然而，当前的AI生命周期指标仅评估孤立的单模型推理，忽视了多智能体执行图，导致网络运营商缺乏基础模型来判断分布式智能体通信的能耗影响。为此，本文提出了agentic-eCAL，通过结合闭式两速单调用能量模型与7层OSI数据传输，推广了AI生命周期的能量成本（eCAL）指标。基于数百个GPU基准配置和多种模型与拓扑的验证，研究了工作流的放置影响，发现智能体间文本传输在5G RAN、城域网和光纤链路中的能耗仅占工作流能耗的0.25%。

🔬 方法详解

问题定义：本文旨在解决现有AI生命周期指标无法有效评估多智能体工作流的能耗问题，导致网络运营商在智能体部署时缺乏科学依据。

核心思路：提出agentic-eCAL指标，通过结合计算密集型预填充和内存密集型解码的闭式两速单调用能量模型，来评估多智能体工作流的能量成本。

技术框架：整体架构包括能量成本评估模块和数据传输模块，前者基于GPU性能基准测试，后者遵循7层OSI模型进行数据传输分析。

关键创新：最重要的技术创新在于将AI生命周期的能量成本推广到多智能体工作流，填补了现有方法在多智能体执行图评估上的空白。

关键设计：在实验中使用了NVIDIA A100和H100 GPU的数百个基准配置，结合16个开放权重模型和8种调度拓扑，确保了模型的广泛适用性和准确性。

🖼️ 关键图片

fig_0
img_1
img_2

📊 实验亮点

实验结果表明，智能体间文本传输在5G RAN、城域网和光纤链路中的能耗仅占工作流总能耗的0.25%。这表明，智能体的能耗主要来自推理和上下文处理，而非文本传输本身，为智能体的部署和优化提供了新的思路。

🎯 应用场景

该研究的潜在应用领域包括智能交通、智能制造和物联网等需要高效能耗管理的边缘计算场景。通过优化智能体的部署位置和通信方式，可以显著降低能耗，提高系统的可持续性和经济性，推动智能体AI技术的广泛应用。

📄 摘要（原文）

As telecommunication networks evolve toward autonomous 5G-Advanced and 6G operations, agentic artificial intelligence (AI) workflows, where large language models (LLMs) execute multi-step reasoning, invoke diagnostic tools, retrieve domain knowledge, and coordinate across agent teams, are increasingly embedded across the edge-cloud continuum. While the biological brain accomplishes complex cognition on an exceptionally modest metabolic power budget of approximately 20W contemporary LLMs are profoundly energy- and memory-intensive, making sustainable lifecycle orchestration a critical operational priority. However, existing AI lifecycle metrics evaluate only isolated, single-model inferences or overlook multi-agent execution graphs entirely. Consequently, network operators lack foundational models to determine whether distributed agent communication incurs meaningful energy costs and where across edge-cloud tiers agent teams should physically reside. To address this gap, we introduce agentic-eCAL, generalizing the Energy Cost of AI Lifecycle (eCAL) metric to directed multi-agent workflows by coupling a closed-form two-rate single-call energy model (compute-bound prefill and memory-bound decode) with 7-layer OSI data transport. Grounded in hundreds of GPU benchmark configurations on NVIDIA A100 and H100, 16 open-weight models and 8 orchestration topologies, we validate components of the metric and study workflow placement implications. Our findings demonstrate that inter-agent text transport incurs 0.25% of workflow energy across 5G RAN, metro, and optical links. Therefore in edge-cloud agent placement the dominant energy cost of distribution is often not the transmission of inter-agent text itself, but the additional inference and context processing induced by that communication.