GPU Passthrough Performance: A Comparison of KVM, Xen, VMWare ESXi, and LXC for CUDA and OpenCL Applications
作者： John Paul Walters , Andrew J. Younge , Dong In Kang , Ke Thia Yao , Mikyung Kang
关键词: CUDA 、 Computer science 、 Cloud computing 、 Hypervisor 、 Operating system 、 Virtualization
摘要: As more scientific workloads are moved into the cloud, need for high performance accelerators increases. Accelerators such as GPUs offer improvements in both and power efficiency over traditional multi-core processors, however, their use cloud has been limited. Today, several common hypervisors support GPU passthrough, but not systematically characterized. In this paper we show that low overhead passthrough is achievable across 4 major two processor microarchitectures. We compare of generations NVIDIA within Xen, VMWare ESXi, KVM hypervisors, also to Linux Containers (LXC). achieves 98 -- 100\% base system's architectures, while Xen achieve 96 99\% systems performance, respectively. addition, describe valuable lessons learned through our analysis share advantages disadvantages each hypervisor/GPU solution.
computer.org
indiana.edu 
isi.edu 参考文章(15)
Giulio Giunta, Raffaele Montella, Giuseppe Agrillo, Giuseppe Coviello, A GPGPU transparent virtualization component for high performance computing clouds european conference on parallel processing. pp. 379- 391 ,(2010) , 10.1007/978-3-642-15277-1_37
Chao-Tung Yang, Hsien-Yi Wang, Wei-Shen Ou, Yu-Tso Liu, Ching-Hsien Hsu, On implementation of GPU virtualization using PCI pass-through ieee international conference on cloud computing technology and science. pp. 711- 716 ,(2012) , 10.1109/CLOUDCOM.2012.6427531
Lin Shi, Hao Chen, Jianhua Sun, Kenli Li, vCUDA: GPU-Accelerated High-Performance Computing in Virtual Machines IEEE Transactions on Computers. ,vol. 61, pp. 804- 816 ,(2012) , 10.1109/TC.2011.112
Jose Duato, Antonio J. Pena, Federico Silla, Rafael Mayo, Enrique S. Quintana-Orti, rCUDA: Reducing the number of GPU-based accelerators in high performance clusters international conference on high performance computing and simulation. pp. 224- 231 ,(2010) , 10.1109/HPCS.2010.5547126
Jiuxing Liu, Evaluating standard-based self-virtualizing devices: A performance study on 10 GbE NICs with SR-IOV support international parallel and distributed processing symposium. pp. 1- 12 ,(2010) , 10.1109/IPDPS.2010.5470365
Steve Plimpton, Fast parallel algorithms for short-range molecular dynamics Journal of Computational Physics. ,vol. 117, pp. 1- 19 ,(1995) , 10.1006/JCPH.1995.1039
Vishakha Gupta, Ada Gavrilovska, Karsten Schwan, Harshvardhan Kharche, Niraj Tolia, Vanish Talwar, Parthasarathy Ranganathan, GViM: GPU-accelerated virtual machines ieee international conference on high performance computing data and analytics. pp. 17- 24 ,(2009) , 10.1145/1519138.1519141
J. Jose, Mingzhe Li, Xiaoyi Lu, K. C. Kandalla, M. D. Arnold, D. K. Panda, SR-IOV support for virtualization on infiniband clusters: early experience ieee acm international symposium cluster cloud and grid computing. pp. 385- 392 ,(2013) , 10.1109/CCGRID.2013.76
Andrew J. Younge, Robert Henschel, James T. Brown, Gregor von Laszewski, Judy Qiu, Geoffrey C. Fox, Analysis of Virtualization Technologies for High Performance Computing Environments 2011 IEEE 4th International Conference on Cloud Computing. pp. 9- 16 ,(2011) , 10.1109/CLOUD.2011.29
Ioannis Kompatsiaris, Andreas Athanasopoulos, Anastasios Dimou, Vasileios Mezaris, GPU acceleration for support vector machines workshop on image analysis for multimedia interactive services. ,(2011)