[1]缪静文,王 召,俞 俊,等.桌面云环境下的高性能 vGPU 计算性能分析[J].计算机技术与发展,2019,29(11):184-189.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 037]
MIAO Jing-wen,WANG Zhao,YU Jun,et al.Performance Analysis of High Performance vGPU in Desktop Cloud Environment[J].,2019,29(11):184-189.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 037]
点击复制
桌面云环境下的高性能 vGPU 计算性能分析(
)
《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]
- 卷:
-
29
- 期数:
-
2019年11期
- 页码:
-
184-189
- 栏目:
-
应用开发研究
- 出版日期:
-
2019-11-10
文章信息/Info
- Title:
-
Performance Analysis of High Performance vGPU in Desktop Cloud Environment
- 文章编号:
-
1673-629X(2019)11-0184-06
- 作者:
-
缪静文; 王 召; 俞 俊; 朱广新; 杨云飞; 程 聪
-
南瑞集团有限公司(国网电力科学研究院有限公司),江苏 南京 211000
- Author(s):
-
MIAO Jing-wen; WANG Zhao; YU Jun; ZHU Guang-xin; YANG Yun-fei; CHENG Cong
-
NARI Group Limited Corporation (State Grid Electric Power Research Institute Limited Corporation),Nanjing 211000,China
-
- 关键词:
-
GPU 虚拟化; 图形加速; 桌面云; 性能分析
- Keywords:
-
GPU virtualization; graphics acceleration; desktop cloud; performance analysis
- 分类号:
-
TP399
- DOI:
-
10. 3969 / j. issn. 1673-629X. 2019. 11. 037
- 摘要:
-
作为一种典型的云计算应用模式,桌面云改变了用户使用计算机的传统方式。 随着桌面云应用的逐渐普及,如何加速图形处理和提高 GPU 利用率,已成为桌面云环境下使用 GPU 的关键问题。 针对该问题,文中利用 vGPU 技术和南瑞桌面云 NCloud-D 搭建原型系统,分析在桌面云环境中使用 vGPU 技术实现 GPU 共享的可行性。 在此基础上有针对性地设计实验,研究对比现有 vGPU 虚拟桌面与物理机的性能,分析不同图形帧缓冲区分配方案下的 vGPU 虚拟桌面性能,分析 vGPU 虚拟桌面多并发的性能。 实验结果表明,vGPU 技术对 GPU 资源有较高的利用效率,桌面云环境下的 vGPU 虚拟桌面与物理机性能相当,基本能满足图形加速处理的实际业务需求。
- Abstract:
-
As a typical cloud computing application model,desktop cloud changes the traditional way users use computers. With the popularity of desktop cloud applications,how to speed up graphics processing and improve GPU utilization have become key issues in the use of vGPU in desktop cloud environment. For this,we use GPU virtualization technology and NCloud-D to build a prototype system for analysis of the feasibility of using GPU sharing in the desktop cloud. On this basis,we design a series of experiments to compare performance between vGPU virtual desktops and the physical machine. In addition,we analyze the performance of vGPU virtual desktops when various amounts of graphics frame buffer assigned and the concurrent performance of vGPU virtual desktops. The experiment shows that the vGPU technology has a high utilization of GPU resources. The vGPU virtual desktop in the desktop cloud has the same performance as the physical machine,which can basically meet the business requirements of graphics acceleration processing.
更新日期/Last Update:
2019-11-10