CarbonShift: Harnessing Grid Carbon Variability for Geo-Distributed Workload Scheduling
Artificial Intelligence and Machine Learning Review · 08.10.2025 · 6 (4) · 18-31
DOI: 10.69987/aimlr.2025.60402 ↗Kurz gefasst
Fachartikel in der Zeitschrift Artificial Intelligence and Machine Learning Review, online erschienen am 08.10.2025. Verfasst von Yanhuan Chen und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Data centers account for 1-1.3% of total U.S. electricity consumption, with carbon emissions escalating rapidly due to artificial intelligence training and cloud computing demands. Current scheduling approaches prioritize performance and cost optimization while largely overlooking the substantial spatio-temporal variability in grid carbon intensity, which can differ by 5-10x across regions and time periods. This paper presents CarbonShift, a carbon-aware scheduling framework that exploits grid carbon intensity variations for geo-distributed workload management. The framework integrates workload energy profiling, LSTM-based carbon intensity forecasting, and deep reinforcement learning-driven optimization to balance carbon reduction with job completion time and data transfer costs. Experimental results demonstrate carbon emission reductions of 42-67% compared to carbon-agnostic scheduling while maintaining acceptable performance trade-offs.
Zitieren
APA
Chen, Y., Chen, Z., & Zou, D. (2025). CarbonShift: Harnessing Grid Carbon Variability for Geo-Distributed Workload Scheduling. Artificial Intelligence and Machine Learning Review, 6(4), 18-31. https://doi.org/10.69987/aimlr.2025.60402
Vancouver
Chen Y, Chen Z, Zou D. CarbonShift: Harnessing Grid Carbon Variability for Geo-Distributed Workload Scheduling. Artificial Intelligence and Machine Learning Review. 2025;6(4):18-31. doi:10.69987/aimlr.2025.60402
BibTeX
@article{Chen2025,
author = {Chen, Yanhuan and Chen, Zijie and Zou, Danbing},
title = {CarbonShift: Harnessing Grid Carbon Variability for Geo-Distributed Workload Scheduling},
journal = {Artificial Intelligence and Machine Learning Review},
year = {2025},
volume = {6},
number = {4},
pages = {18--31},
doi = {10.69987/aimlr.2025.60402}
}
EUVIMED