跳到正文
arXiv cs.AI· Xin Qin, Dawei Qiu, Fei Teng·· 2 小时前AI 评分32

免求解器的 AI 数据中心任务成本感知在线调度方法

Solver-Free and Cost-Aware Online Scheduling of AI Data Center Tasks

AI 导读

一篇 arXiv 论文提出受对偶分解启发的免求解器在线优化规则,为 AI 数据中心持续到达的计算任务在到达时分配起始时段,将灵活任务移向低价时段同时保证高优先级任务按计划执行。在 3,456 张 A100 GPU、1,343 个任务的 GPU 负载轨迹仿真中,该方法在保持 GPU 和 CPU 用量不超容量的同时实现更低成本,且相比多时间尺度优化基准大幅减少计算时间。

正文 · 原文

View PDF HTML (experimental)

Abstract:AI data centers receive computing tasks continuously, whereas solver-based optimization takes time to run and electricity is priced per settlement period. This paper bridges this timing gap through a solver-free online optimization rule inspired by dual decomposition. The rule assigns each task a start period upon arrival to optimize energy cost subject to operational constraints, shifting flexible tasks toward low-price periods while keeping high-priority tasks on schedule. Simulations on the A100 GPU trace (3,456 GPUs, 1,343 tasks) show that the proposed rule keeps GPU and CPU use within capacity while scheduling tasks at low cost. Compared with a multi-timescale optimization benchmark, it achieves a lower cost and, because it needs no solver, much less computation time.
Subjects: Systems and Control (eess.SY); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.22239 [eess.SY]
  (or arXiv:2611.22239v1 [eess.SY] for this version)
  https://doi.org/10.48550/arXiv.2611.22239

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Xin Qin [view email]
[v1] Tue, 17 Nov 2026 05:39:35 UTC (817 KB)

来源:arXiv cs.AI · arxiv.org