@inproceedings{chang-etal-2026-spiderflow,
    title = "{S}pider{F}low: Efficient Topology-Aware Scheduling for {LLM} Training Across Decentralized {GPU} Clusters",
    author = "Chang, Zihan  and
      He, Shuibing  and
      Zhou, Bo  and
      Xiao, Sheng  and
      Yang, Siling  and
      Wang, Rui  and
      Pan, Zhe",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.619/",
    pages = "13603--13615",
    ISBN = "979-8-89176-390-6",
    abstract = "In response to the increasing demand for largescale machine learning training jobs, many organizations have deployed GPU clusters across geographically distributed regions. However, existing ILP- or genetic-based cross-cluster training approaches largely overlook the topology of decentralized clusters, lacking both topologyaware task scheduling mechanisms and automated model parallelization strategies. As a result, naively applying these optimization-based methods in cross-cluster settings leads to prohibitive scheduling overhead, due to the drastically enlarged search space induced by complex inter-cluster topologies. To address these challenges, we propose SpiderFlow, a topologyaware scheduling system specifically designed for decentralized GPU clusters. We formulate cross-cluster task scheduling as a graph optimization problem and introduce SpinSearch, a low-overhead topology-aware scheduling algorithm. In addition, for automated model parallelization, we propose TPA, a two-level scheduling framework that combines heuristic methods at the inter-cluster level with ILP-based optimization within clusters, effectively reducing the search space while maintaining high training throughput with substantially lower scheduling overhead. We evaluate SpiderFlow on a physical platform comprising 8 decentralized clusters, as well as on a simulation platform with up to 64 decentralized clusters. Experimental results demonstrate that SpiderFlow reduces job completion time (JCT) by 1.2-1.3{\texttimes}, improves throughput by 1.12-1.25{\texttimes}, and reduces scheduling overhead by 20-90{\texttimes} on average compared to state-of-the-art scheduling systems."
}