Performance Isolation and Semantic Determinism in Efficient GPU Spatial Sharing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Zhenyuan, Zheng, Wenxin, Li, Mingyu, Chen, Haibo
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910098773770240
author Yang, Zhenyuan
Zheng, Wenxin
Li, Mingyu
Chen, Haibo
author_facet Yang, Zhenyuan
Zheng, Wenxin
Li, Mingyu
Chen, Haibo
contents Existing GPU spatial sharing systems face a three-way tradeoff: resource utilization, performance isolation, and semantic determinism. Hardware partitioning suffers from hardware under-utilization. Hardware multiplexing fails to avoid performance interference. Recently proposed software-based GPU kernel slicing reshapes floating-point reduction orders, destroying semantic determinism and inducing catastrophic token drift in generative models. We present CoGPU, a transparent spatial sharing system that resolves this trilemma. CoGPU introduces \emph{GPU coroutine}, a novel abstraction that enables logical-to-physical resource decoupling. By dynamically mapping immutable virtual contexts to mutable physical resource via lightweight cooperative migration, CoGPU enables extensible, workload-aware scheduling without altering kernel semantics. Evaluations demonstrate CoGPU simultaneously achieves high utilization, strong isolation, and absolute semantic determinism (guaranteeing zero token mismatch). In multi-tenant co-location, it improves training throughput by up to 79.2\% over temporal sharing and reduces P99 inference tail latency by 15.1\%. Its pluggable architecture supports custom policies; compared to the default policy, a \textsc{TPOT-FIRST} policy further reduces SLO violations by 21.2\% under dynamic traffic.
format Preprint
id arxiv_https___arxiv_org_abs_2603_15042
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Performance Isolation and Semantic Determinism in Efficient GPU Spatial Sharing
Yang, Zhenyuan
Zheng, Wenxin
Li, Mingyu
Chen, Haibo
Distributed, Parallel, and Cluster Computing
Operating Systems
Existing GPU spatial sharing systems face a three-way tradeoff: resource utilization, performance isolation, and semantic determinism. Hardware partitioning suffers from hardware under-utilization. Hardware multiplexing fails to avoid performance interference. Recently proposed software-based GPU kernel slicing reshapes floating-point reduction orders, destroying semantic determinism and inducing catastrophic token drift in generative models. We present CoGPU, a transparent spatial sharing system that resolves this trilemma. CoGPU introduces \emph{GPU coroutine}, a novel abstraction that enables logical-to-physical resource decoupling. By dynamically mapping immutable virtual contexts to mutable physical resource via lightweight cooperative migration, CoGPU enables extensible, workload-aware scheduling without altering kernel semantics. Evaluations demonstrate CoGPU simultaneously achieves high utilization, strong isolation, and absolute semantic determinism (guaranteeing zero token mismatch). In multi-tenant co-location, it improves training throughput by up to 79.2\% over temporal sharing and reduces P99 inference tail latency by 15.1\%. Its pluggable architecture supports custom policies; compared to the default policy, a \textsc{TPOT-FIRST} policy further reduces SLO violations by 21.2\% under dynamic traffic.
title Performance Isolation and Semantic Determinism in Efficient GPU Spatial Sharing
topic Distributed, Parallel, and Cluster Computing
Operating Systems
url https://arxiv.org/abs/2603.15042