Download presentation
Presentation is loading. Please wait.
Published byDustin Lester Modified over 7 years ago
1
GPU Virtualization at Xen Hypervisor SungJu Yoo, YoungHun Park
Distributed and Cloud Computing Lab ( Dept. of Computer Science and Engineering Sogang University, Seoul, KOREA Paper PT
2
What is GPU Virtualization?
3
What is GPU Virtualization?
4
GPU Virtualization Classification
Front-end virtualization Device emulation API forwarding Back-end virtualization (Fixed) pass-through Mediated pass-through
5
API remoting API remoting
6
API remoting API remoting(forwarding)
rCUDA, DS-CUDA, vCUDA, GViM, gVirtuS, GridCuda [1, 2, 3, 4, 5, 6] Currently, only rCUDA support latest CUDA SDK 8.0 and ongoing project with sponsors. But it is not open source. Device Emulation It has great complexity and extremely low performance [7] [1] Duato, José, et al. "rCUDA: Reducing the number of GPU-based accelerators in high performance clusters." High Performance Computing and Simulation (HPCS), 2010 International Conference on. IEEE, 2010. [2] Oikawa, Minoru, et al. "DS-CUDA: a middleware to use many GPUs in the cloud environment." High Performance Computing, Networking, Storage and Analysis (SCC), 2012 SC Companion:. IEEE, 2012. [3] Shi, Lin, et al. "vCUDA: GPU-accelerated high-performance computing in virtual machines." IEEE Transactions on Computers 61.6 (2012): [4] Gupta, Vishakha, et al. "GViM: GPU-accelerated virtual machines." Proceedings of the 3rd ACM Workshop on System-level Virtualization for High Performance Computing. ACM, 2009. [5] Giunta, Giulio, et al. "A GPGPU transparent virtualization component for high performance computing clouds." European Conference on Parallel Processing. Springer Berlin Heidelberg, 2010. [6] Liang, Tyng-Yeu, and Yu-Wei Chang. "GridCuda: a grid-enabled CUDA programming toolkit." Advanced Information Networking and Applications (WAINA), 2011 IEEE Workshops of International Conference on. IEEE, 2011. [7] Bellard, Fabrice. "QEMU, a fast and portable dynamic translator." USENIX Annual Technical Conference, FREENIX Track
7
Pass-through Pass-through
8
Full GPU Virtualization
9
Full GPU virtualization
Pass-through Amazon EC2, Intel VT-d [8, 9] Virtualization of hypervisor Intel GVT-G technology (XENGT, KVMGT) [10] Full GPU virtualization in Xen hypervisor It only support Intel integrated GPU. GPUvm [11] Para GPU virtualization in Xen hypervisor It support NVIDIA GPU and CUDA [8] AMAZON.COM. Amazon Elastic Compute Cloud (Amazon EC2) [9] Abramson, Darren, et al. "Intel Virtualization Technology for Directed I/O." Intel technology journal 10.3 (2006). [10] Tian, Kun, Yaozu Dong, and David Cowperthwaite. "A Full GPU Virtualization Solution with Mediated Pass-Through." USENIX Annual Technical Conference [11] Suzuki, Yusuke, et al. "GPUvm: Why not virtualizing GPUs at the hypervisor?." USENIX Annual Technical Conference
10
Full GPU virtualization
Pass-through Amazon EC2, Intel VT-d [8, 9] Virtualization of hypervisor Intel GVT-G technology (XENGT, KVMGT) [10] Full GPU virtualization in Xen hypervisor It only support Intel integrated GPU. GPUvm [11] Para GPU virtualization in Xen hypervisor It support NVIDIA GPU and CUDA [8] AMAZON.COM. Amazon Elastic Compute Cloud (Amazon EC2) [9] Abramson, Darren, et al. "Intel Virtualization Technology for Directed I/O." Intel technology journal 10.3 (2006). [10] Tian, Kun, Yaozu Dong, and David Cowperthwaite. "A Full GPU Virtualization Solution with Mediated Pass-Through." USENIX Annual Technical Conference [11] Suzuki, Yusuke, et al. "GPUvm: Why not virtualizing GPUs at the hypervisor?." USENIX Annual Technical Conference
11
Q & A
Similar presentations
© 2025 SlidePlayer.com Inc.
All rights reserved.