Article Details

Title · Authors · Abstract · Keywords · DOI · References

Dynamic Resource Partitioning and Lightweight Scheduling Method for vGPU

CHEN Xudong, WANG Linqiang, SHEN Xuedong, LI Mengxiang, CHEN Mo

(China Academy of Launch Vehicle Technology, Beijing 100076, China)

Section：Computer Technology · Vol. 2026, Issue 12 · Pages：87-90 · 2096-4706（2026）12-0087-04 · CLC number：TP338.6；TP316.8 · Document code：A · DOI：10.19850/j.cnki.2096-4706.2026.12.017

Abstract

GPU is a high-performance computing resource urgently needed in fields such as simulation and design, artificial intelligence, etc. GPU virtualization improves GPU utilization, but traditional static partitioning suffers from low resource utilization and poor scenario adaptability due to fixed allocation strategies. For the application scenario of virtual design desktops in small businesses, this paper proposes a lightweight vGPU resource scheduling scheme with “minimum initial configuration plus dynamic incremental adjustment.” The scheme consists of three core modules: resource collection, dynamic partitioning, and quota adjustment, to achieve fine-grained vGPU resource configuration. Experimental results show that, under equivalent design workload conditions, the proposed scheme improves average vGPU utilization from 30% to 70%, and increases the number of virtual design desktops supported by a single 8 GB VRAM GPU from 2 to 5, effectively addressing the coexistence of GPU resource shortage and waste in small businesses, and reduces the technical barrier of virtualized resource scheduling.

Keywords

GPU virtualization; dynamic partitioning; resource scheduling; workload monitoring; virtual desktop

References

  1. [1] Li Yurong, Liang Guicai Implementation of an Artificial Intelligence Teaching Platform Based on GPU Virtualization [J]. Information Technology, 2024 (8): 71-78
  2. [2] Ning Shiyong Research on Energy saving Scheduling Algorithm for Virtual Cloud Computing Data Center Resources [J]. Computer Application Research, 2021, 38 (4): 1088-1091
  3. [3] Chen Qinhua Discussion on the Construction of Enterprise Cloud Desktop System [J]. Network Security Technology and Application, 2020 (4): 93-95
  4. [4] Han Dongliang, He Xiaochen, Li Shaofei Application of vGPU based desktop virtualization in 3D design [J]. Electronic Industry Special Equipment, 2021, 50 (1): 52-56
  5. [5] Fu Maozhong, Hu Haiyang, Li Zhongjin Dynamic Resource Scheduling Method for GPU Clusters [J]. Computer Research and Development, 2023, 60 (6): 1308-1321
  6. [6] Zhang Yunpeng, Zang Yanan, Tian Qiuyu Research on GPU Cloud Resource Reservation based Intelligent Scheduling Technology [J]. Ship Electronic Engineering, 2025, 45 (2): 119-123
  7. [7] Wang Zhuang, Wang Pinghui, Wang Bincheng, etc GPU shared scheduling system under deep learning container cloud platform [J]. Computer Science, 2023, 50 (6): 86-91
  8. [8] Tang Xiaochun, Zhu Ziyu, Mao Anqi, etc Research on Scheduling Algorithms for Data Intensive Jobs on GPU Clusters [J]. Journal of Software, 2022, 33 (12): 4429-4451
  9. [9] Qian Min Research on Resource Scheduling Methods for Data Centers under the Background of Cloud Computing [J]. Information Technology, 2022 (5): 188-192
  10. [10] Zhang Wancai, Zhang Nan, Yang Wenqing, etc Architecture Design, Key Technologies, and Application Research of GPU Heterogeneous Resource Pool Platform Based on Virtualization [J]. Telecommunications Science, 2024, 40 (9): 162-175

About the authors

Chen Xudong (1979.11-), male, Han ethnicity, born in Yuncheng, Shanxi, is a senior engineer and a master's student. His research interests include cloud platforms and virtualization technology, as well as video conferencing.