一句话：提出细粒度线程块级 GPU 内核调度机制 Tally，实现非侵入式 GPU 共享与稳健性能隔离。
GPU 集群中利用率不足会导致队列延长和运营成本上升，GPU 共享是提升利用率的有效途径，但现有机制存在集成成本高、性能隔离不足与应用兼容性有限等问题。本文提出 Tally，一种非侵入式 GPU 共享机制，其核心是细粒度线程块级 GPU 内核调度策略，可在多个工作负载并发执行时有效抑制相互干扰，同时保持广泛兼容性。作者在多种工作负载上评估 Tally：当高优先级推理任务与尽力而为的训练任务并发执行时，Tally 对推理任务 99 分位延迟平均仅带来 7.2% 开销，而当前先进系统 TGS 的开销为 188.9%；同时 Tally 达到 TGS 系统吞吐的 80% 以上。
GPU underutilization is a significant concern in many production deep learning clusters, leading to prolonged job queues and increased operational expenses. A promising solution to this inefficiency is GPU sharing, which improves resource utilization by allowing multiple workloads to execute concurrently on a single GPU. However, deploying GPU sharing in production settings faces critical obstacles due to the limitations of existing mechanisms, including high integration costs, inadequate performance isolation, and limited application compatibility. To address these issues, we introduce Tally, a non-intrusive GPU sharing mechanism that provides robust performance isolation and comprehensive workload compatibility. The key to Tally's robust performance isolation capability lies in its fine-grained thread-block-level GPU kernel scheduling strategy, which allows the system to effectively mitigate interference caused by workload co-execution. We evaluate Tally on a diverse range of workloads and show that it incurs an average overhead of only 7.2% on the 99th-percentile latency of high-priority inference tasks when executed concurrently with best-effort training workloads, compared to 188.9% overhead exhibited by the state-of-the-art GPU sharing systems like TGS, while achieving over 80% of TGS's system throughput.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{abs-2410-07381,
title = {Tally: Non-Intrusive Performance Isolation for Concurrent Deep Learning Workloads},
author = {Wei Zhao and Anand Jayarajan and Gennady Pekhimenko},
booktitle = {ASPLOS},
year = {2025},
doi = {10.1145/3669940.3707282}
}