[论文解读] On the Opportunities of Green Computing: A Survey

Y. Zhou, Xiujing Lin+39 more|arXiv (Cornell University)|Nov 1, 2023

Green IT and Sustainability被引用 9

一句话总结

本综述提出一个 Green Computing 框架，以在 AI 性能、资源使用和环境影响之间取得平衡，并回顾在模型设计、训练、推理和系统方面的技术。

ABSTRACT

Artificial Intelligence (AI) has achieved significant advancements in technology and research with the development over several decades, and is widely used in many areas including computing vision, natural language processing, time-series analysis, speech synthesis, etc. During the age of deep learning, especially with the arise of Large Language Models, a large majority of researchers' attention is paid on pursuing new state-of-the-art (SOTA) results, resulting in ever increasing of model size and computational complexity. The needs for high computing power brings higher carbon emission and undermines research fairness by preventing small or medium-sized research institutions and companies with limited funding in participating in research. To tackle the challenges of computing resources and environmental impact of AI, Green Computing has become a hot research topic. In this survey, we give a systematic overview of the technologies used in Green Computing. We propose the framework of Green Computing and devide it into four key components: (1) Measures of Greenness, (2) Energy-Efficient AI, (3) Energy-Efficient Computing Systems and (4) AI Use Cases for Sustainability. For each components, we discuss the research progress made and the commonly used techniques to optimize the AI efficiency. We conclude that this new research direction has the potential to address the conflicts between resource constraints and AI development. We encourage more researchers to put attention on this direction and make AI more environmental friendly.

研究动机与目标

• 推动建立 Green Computing 的必要性，以解决 AI 发展中的资源和环境问题。
• 提出一个 four-component Green Computing 框架（Measures of Greenness、Energy-Efficient AI、Energy-Efficient Computing Systems、AI Use Cases for Sustainability）。
• 综述在设计、训练、推理和部署阶段提高 AI 效率的当前进展与技术。
• 突出促进环境友好型 AI 研究与实践的机会、挑战与未来方向。

提出的方法

• 使用四个组件定义 Green Computing 框架并解释它们的作用。
• 回顾 Measures of Greenness，包括运行时间、模型大小、FLOPs、硬件功率、能量和碳排放。
• 调研高效能的模型设计、训练和推理技术（例如紧凑模块、NAS、剪枝、量化、蒸馏、早期退出）。
• 讨论部署时的高效能计算系统与数据管理实践。
• 审视 Green Large Language Models 与 AI for sustainability use cases。
• 确定用于跟踪 greenness 与碳足迹的工具包与指标。

实验结果

研究问题

• RQ1 哪些因素和指标最能量化 AI 系统的 greenness？
• RQ2 哪些技术在不牺牲性能的前提下优化模型设计、训练和推理的能效？
• RQ3 如何将 Green Computing 集成到 AI 系统和工作流程中，以促进可持续性与公平性？
• RQ4 在研究与行业中采用 Green Computing 的实际机会与挑战有哪些？

主要发现

• Green Computing 被框定为一个 four-component 框架，包含 measures、energy-efficient AI、systems 和 sustainability use cases。
• 常见的 greenness 指标包括运行时间、模型大小、FLOPs、硬件功率、能量和碳排放。
• 在高效能模型设计与训练方面存在广泛的技术，以及推理优化如剪枝、量化、蒸馏和早期退出。
• 工具包和框架（例如跟踪 FLOPs、能量和碳）支持衡量 AI 的 greenness，使基准测试透明化。
• Green Computing 提供降低成本、实现边缘部署，以及通过降低资源门槛来促进研究公平性的机会。
• 本综述强调 Green Computing 在平衡 AI 进展与环境及社会考量方面的潜力。