LHRS-LTW: A Load-Aware Hybrid-Cloud Resource Scheduling Framework for Large-Scale Training Workloads
IEEE Transactions on Parallel and Distributed Systems · 2026-08 · 37 (8) · 1917-1934
DOI: 10.1109/tpds.2026.3704297 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Parallel and Distributed Systems, 2026-08. Verfasst von Ao Wei und 6 weiteren Autor:innen. Die Arbeit nennt 65 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Scientific and Technological Innovation Platform Research Project of Guizhou Province, Key Project of Guizhou Basic Research Program.
Abstract (Originalsprache)
To address the increasing computational demands driven by the development of large-scale pretrained AI models, hybrid-cloud environments have emerged as an effective paradigm for elastically extending computational resources while balancing the cloud cost and job completion time (JCT). Recent advanced schedulers aim to dynamically assign waiting budgets according to job-specific computational loads. However, these schedulers often rely on deterministic, single-shot mathematical models that cannot be adequately adapted to real-time fluctuations in the cluster state. Under bursty workloads, the absence of multi-round, feedback-driven load awareness can allow long-tail jobs to monopolize scarce resources, thereby increasing tail latency and cloud costs. To address this issue, we propose LHRS-LTW, a load-aware hybrid-cloud resource scheduling framework for large-scale training workloads. The key innovations of our framework are as follows: $i$ i ) Multi-round cardinality-aware resource profiling algorithms (MCRPs) are established, which introduce cardinal queries into hybrid-cloud scheduling. Through a three-phase iterative process of load sensitivity determination, starvation detection, and trend analysis, this mechanism dynamically generates resource-matching factors and significantly increases job-node affinity and scheduling flexibility under bursty workloads. $ii$ i i ) Load-aware adaptive deep reinforcement learning for dynamic workloads (LAADRL-DW) is developed. By introducing dynamic learning factors, online normalization, and asynchronous multithreaded parallel updates, LAADRL-DW achieves real-time adaptive optimization of scheduling strategies and responds effectively to load fluctuations and heterogeneous resource contention. $iii$ i i i ) An intelligent offloading system (IOS) is designed. This module relies on a hierarchical offloading strategy that accounts for the queue length, resource saturation, hardware affinity, and cost efficiency. By implementing dynamic threshold triggering, affinity-based scoring, overload prediction, and cost-efficiency balancing, the IOS ensures local execution of high-priority and large-scale tasks while flexibly offloading inefficient or CPU-intensive tasks to the cloud. The experimental results demonstrate that, under moderate to low workload conditions, LHRS-LTW achieves superior cost savings, providing additional savings of up to 40.56%. Under highly bursty workload scenarios, the collaborative interaction among the proposed modules achieves cost savings of up to 17% and increases cluster utilizationto 99.2%.
Förderung
- National Natural Science Foundation of China
- Scientific and Technological Innovation Platform Research Project of Guizhou Province
- Key Project of Guizhou Basic Research Program
- Science and Technology Research Project of Guizhou Provincial Department of Education
- Guizhou Provincial Key Technology R&D Program
- Guizhou University Science and Technology Innovation Team
- Guizhou University Basic Research Fund
Zitieren
APA
Wei, A., Yang, J., Pang, P., Sun, S., Ruan, X., Chen, Y., & Guo, M. (2026). LHRS-LTW: A Load-Aware Hybrid-Cloud Resource Scheduling Framework for Large-Scale Training Workloads. IEEE Transactions on Parallel and Distributed Systems, 37(8), 1917-1934. https://doi.org/10.1109/tpds.2026.3704297
Vancouver
Wei A, Yang J, Pang P, Sun S, Ruan X, Chen Y, et al. LHRS-LTW: A Load-Aware Hybrid-Cloud Resource Scheduling Framework for Large-Scale Training Workloads. IEEE Transactions on Parallel and Distributed Systems. 2026;37(8):1917-1934. doi:10.1109/tpds.2026.3704297
BibTeX
@article{Wei2026,
author = {Wei, Ao and Yang, Jing and Pang, Pu and Sun, Shixuan and Ruan, Xiaoli and Chen, Yuling and Guo, Minyi},
title = {LHRS-LTW: A Load-Aware Hybrid-Cloud Resource Scheduling Framework for Large-Scale Training Workloads},
journal = {IEEE Transactions on Parallel and Distributed Systems},
year = {2026},
volume = {37},
number = {8},
pages = {1917--1934},
doi = {10.1109/tpds.2026.3704297}
}
EUVIMED