EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

GreenScheduler: Coordinated Two-Tier Energy Optimization for Disaggregated LLM Serving

Waled Milad Abulgasem Alashheb, Mabruka Khlifa Ali Karkeb, Sabria AbdulGader Ali Elmusrati, Sumia Abdussalam Milad Elagtel

مجلة العلوم الشاملة · 01.03.2026 · 10 (39) · 3804-3811

DOI: 10.65405/cn221159 ↗

Kurz gefasst

Fachartikel in der Zeitschrift مجلة العلوم الشاملة, online erschienen am 01.03.2026. Verfasst von Waled Milad Abulgasem Alashheb und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Large Language Model (LLM) inference has become a dominant consumer of en- ergy in modern AI data centers, often accounting for over 90% of total operational power [1].Recent architectural shifts toward prefill/decode disaggregation have improved perfor- mance but created complex energy optimization challenges. This paper introduces Green- Scheduler, a novel two-tier framework designed to jointly optimize GPU placement and Dynamic Voltage and Frequency Scaling (DVFS) in disaggregated environments. Tier 1 performs coarse-grained (minute-scale) phase-aware provisioning using predictive work- load modeling, while Tier 2 executes fine-grained (millisecond-scale) frequency control. For the compute-bound prefill stage, GreenScheduler employs Model Predictive Control (MPC) to manage queue dynamics; for the memory-bound decode stage, it utilizes a lightweight slack-aware adaptation mechanism. Evaluations using production Azure traces [2] on an H100 cluster demonstrate that GreenScheduler achieves significant energy reduc- tion in both decode and prefill pools compared to performance-optimized baselines like DistServe [3], while strictly maintaining Time to First Token (TTFT) and Time Per Output Token (TPOT) Service-Level Objectives (SLOs).

Zitieren

APA

Alashheb, W. M. A., Karkeb, M. K. A., Elmusrati, S. A. A., & Elagtel, S. A. M. (2026). GreenScheduler: Coordinated Two-Tier Energy Optimization for Disaggregated LLM Serving. مجلة العلوم الشاملة, 10(39), 3804-3811. https://doi.org/10.65405/cn221159

Vancouver

Alashheb WMA, Karkeb MKA, Elmusrati SAA, Elagtel SAM. GreenScheduler: Coordinated Two-Tier Energy Optimization for Disaggregated LLM Serving. مجلة العلوم الشاملة. 2026;10(39):3804-3811. doi:10.65405/cn221159

BibTeX
@article{Alashheb2026,
  author = {Alashheb, Waled Milad Abulgasem and Karkeb, Mabruka Khlifa Ali and Elmusrati, Sabria AbdulGader Ali and Elagtel, Sumia Abdussalam Milad},
  title = {GreenScheduler: Coordinated Two-Tier Energy Optimization for Disaggregated LLM Serving},
  journal = {مجلة العلوم الشاملة},
  year = {2026},
  volume = {10},
  number = {39},
  pages = {3804--3811},
  doi = {10.65405/cn221159}
}
Kostenlos starten →