Page 59 - 《软件学报》2020年第10期

李鼎基 等:基于跨虚拟机零下陷通信的加速器虚拟化框架 3035

制化、易于更新的加速器虚拟化系统提供了支持.Wormhole 加速器虚拟化框架以 API 转发为基础,以虚拟机为
隔离保护域,创新性地提出了被动式服务端虚拟机的抽象以及跨虚拟机快速代理执行的各项技术,在保证了用
户间强隔离性的前提下,实现了高硬件资源利用率、低虚拟化开销的加速器虚拟化,并在主流的 QEMU/KVM 平
台上实现了针对 NVIDIA GPU 的原型系统.测试结果表明,Wormhole 可以方便地部署在消费级服务器上,对比
扩展优化后的 GPU 虚拟化代表性方案 GVirtuS,有大幅度的性能提升,验证了本加速器虚拟化框架的有效性.

References:
[1] Zhang ZK, Pang WG, Xie WJ, Lü MS, Wang Y. A survey of deep learning research for real-time applications. Ruan Jian Xue
Bao/Journal of Software, 2020,31(9):2654−2677 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5946.htm
[doi: 10.13328/j.cnki.jos.005946]
[2] Jouppi NP, Young C, Patil N, et al. In-datacenter performance analysis of a tensor processing unit. In: Proc. of the 44th Annual
Int’l Symp. on Computer Architecture (ISCA 2017). New York: Association for Computing Machinery, 2017. 1−12. [doi: 10.1145/
3079856.3080246]
[3] Zhang XL, Yang JH, Sun XQ, Wu JP. Survey of geo-distributed cloud research progress. Ruan Jian Xue Bao/Journal of Software,
2018,29(7):2116−2132 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5555.htm [doi: 10.13328/j.cnki.jos.
005555]
[4] Gao Q, Zhang FL, Wang RJ, Zhou F. Trajectory big data: A review of key technologies in data processing. Ruan Jian Xue
Bao/Journal of Software, 2017,28(4):959−992 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5143.htm [doi:
10.13328/j.cnki.jos.005143]
[5] Intel platform hardware support for I/O virtualization. 2006. http://www.intel.com
[6] Herrera A. NVIDIA GRID: Graphics accelerated VDI with the visual performance of a workstation. White Paper, NVIDIA Corp.,
2014. 1−18.
[7] Tian K, Dong YZ, Cowperthwaite D. A full GPU virtualization solution with mediated pass-through. In: Proc. of the 2014 USENIX
Conf. on USENIX Annual Technical Conf. (USENIX ATC 2014). USENIX Association, 2014. 121−132.
[8] GRID Virtual GPU User Guide. 2020. https://docs.nvidia.com/grid/4.3/grid-vgpu-user-guide/index.html
[9] Duato J, Peña AJ, Silla F, Mayo R, Quintana-Ortí ES. rCUDA: Reducing the number of GPU-based accelerators in high
performance clusters. In: Proc. of the Int’l Conf. on High Performance Computing & Simulation. Caen, 2010. 224−231. [doi: 10.
1109/HPCS.2010.5547126]
[10] Montella R, Giunta G, Laccetti G, Lapegna M, Palmieri C, Ferraro C, Pelliccia V, Hong C-H, Spence I, Nikolopoulos DS. On the
virtualization of CUDA based GPU remoting on ARM and X86 machines in the GVirtuS framework. Int’l Journal of Parallel
Programming, 2017,45(5):1142−1163. [doi: 10.1007/s10766-016-0462-1]
[11] Armand F, Gien M, Maigné G, Mardinian G. Shared device driver model for virtualized mobile handsets. In: Proc. of the 1st
Workshop on Virtualization in Mobile Computing (MobiVirt 2008). New York: Association for Computing Machinery, 2008.
12−16. [doi: 10.1145/1622103.1622104]
[12] Zhang YQ, Wang XF, Liu XF, Liu L. Survey on cloud computing security. Ruan Jian Xue Bao/Journal of Software, 2016,27(6):
1328−1348 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5004.htm [doi: 10.13328/j.cnki.jos.005004]
[13] Yu QQ, Dong MK, Chen HB. Memory-assisted synchronization mechanism for hardware transactions in a virtual environment. Ji
Suan Ji Ke Xue Yu Tan Suo/Journal of Frontiers of Computer Science and Technology, 2017,11(9):1429−1438 (in Chinese with
English abstract).
[14] Wu S, Wang K, Jin H. Research status and prospect of operating system virtualization. Ji Suan Ji Yan Jiu Yu Fa Zhan/Computer
Technology and Development, 2019,56(1):58−68 (in Chinese with English abstract).
[15] Liu YT, Chen HB. Virtualization security: Opportunities, challenges and future. Wang Luo Yu Xin Xi An Quan Xue Bao/Chinese
Journal of Network and Information Security, 2016,2(10):17−28 (in Chinese with English abstract).
[16] Huang X, Deng L, Sun H, Zeng QK. Hardware virtualization-based secure and efficient kernel monitoring model. Ruan Jian Xue
Bao/Journal of Software, 2016,27(2):481−494 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/4866.htm [doi:
10.13328/j.cnki.jos.004866]