臺灣博碩士論文加值系統

現今GPU 的運算能力遠高於CPU。在高效能計算中GPU 日趨重要， 但是若要使用到 GPU 的運算能力，使用者往往需要花費高昂成本購買GPU 晶片， 並不一定符合預算。而在雲端運算的概念被提出後， 使用者可以花費相對較少的成本，使用到相同功能的機器，因此使用者會期望雲端運算業者提供支援虛擬GPU機器的服務。
目前提供承租 GPU 虛擬機器的業者多是採用一台虛擬機器綁定一個GPU 晶片的政策，或是採用傳送檔案到遠端機器處理後再回傳的方式執行GPU程式，並不能做到單一晶片的分時共享。本論文的目的在於設計一套能夠創建GPU虛擬機器的雲端運算系統，使系統可以利用NVidia Fermi 架構， 來同時執行好幾個來自不同虛擬機器的程式，並且達成裝置分時共享的目的。

Nowadays a graphic processing unit (GPU) delivers much better performance than CPU does. As a result GPU is becoming increasingly important in high performance computing (HPC) because of its tremendous computing power. At the same time cloud computing is becoming increasingly popular, and HPC community will expect cloud computing companies to provide virtual GPU service, just like virtual CPU, virtual disk, and virtualized network they have been providing. This business opportunity means computing GPU will be more economical because users can spend relatively less to rent GPUs to fit their computing needs, rather than buying them.
The current practice of virtual GPU rental service is to bind a GPU to a virtual machine statically. The problem of this approach is that virtual machines cannot share a GPU. This is against the "multiple tenancy" principle of cloud computing. The goal of this paper is to design a cloud computing system that can combine the CUDA program from every virtual machine and execute them concurrently, which support the concept of GPU "time-sharing" principle of cloud computing, which is crucial to providing an economic computing service.
This paper describes a method of using NVidia Fermi Architecture GPU for GPU virtualization. The key idea in our design is that NVidia Fermi architecture GPU can support concurrent kernel execution, which can allows 16 CUDA kernels concurrent execution. We virtualize GPU by collecting GPU programs from different virtual machines into a single virtual machine, where the programs will be run by the GPU. Our approach reduces the compilation and execution time of combined programs, and also reduce the average waiting time of each CUDA program from different virtual machines.
We conduct experiments to evaluate the efficiency of our GPU virtualization. Preliminary results are satisfactory.

Certification 1
Acknowledgement i
Chinese Abstract ii
Abstract iii
1 Introduction 1
2 Related Work 4
3 System Architecture 8
3.1 System Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1 Listener . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.2 Combiner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.3 Executor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 System Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4 System Implementation 11
4.1 System Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Listener . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 Combiner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3.1 Concurrent Kernel Execution . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3.2 Choosing Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3.3 Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.4 Executor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5 Experimentation Performance 14
5.1 Experiment Surrounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.2 Parallel Execution Overhead . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.3 Dispatch System Overhead . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.4 Average Waiting Time of CUDA Programs . . . . . . . . . . . . . . . . . . . . 17
6 Conclusion and FutureWork 20
Bibliography 21

[1] Amazon elastic compute cloud. http://aws.amazon.com/ec2/.
[2] Google cloud platform. http://cloud.google.com/.
[3] Xen. http://xen.org.
[4] Kvm. http://www.linux-kvm.org/.
[5] Amd radeon hd 6990 graphics. http://tinyurl.com/69qxshp.
[6] Intel i7-980 xe. http://tinyurl.com/86mmt37.
[7] J.A. Anderson, C.D. Lorenz, and A. Travesset. General purpose molecular dynamics simulations fully implemented on graphics processing units. In Journal of Computational Physics, 2008, pages 5342–5359, February 2008.
[8] G. Chen, G. Li, S. Pei, and B. Wu. Gpgpu supported cooperative acceleration in molecular dynamics. In Computer Supported Cooperative Work in Design (CSCWD), 2009. 13th International Conference on, pages 113–118, April 2009.
[9] Vincent A. Voelz, Gregory R. Bowman, Kyle Beauchamp, and Vijay S. Pande. Molecular simulation of ab initio protein folding for a millisecond folder ntl9(1-39). Journal of the American Chemical Society, 132(5):1526–1528, 2010. PMID: 20070076.
[10] Gpgpu. http://gpgpu.org/.
[11] Nvidia. http://www.nvidia.com/.
[12] Ibm. http://www.ibm.com/.
[13] Intel. http://www.intel.com/.
[14] Amd. http://www.amd.com/.
[15] Cuda. http://www.nvidia.com/content/cuda/cuda-toolkit.html.
[16] Opencl. http://www.khronos.org/opencl/.
[17] Amazon high performance computing on cloud. http://aws.amazon.com/hpc-applications/.
[18] Hoopoe. http://www.hoopoe-cloud.com/.
[19] zillians. http://www.zillians.com/.
[20] Nvidia fermi architecture. http://tinyurl.com/6vdsl4q.
[21] I. Buck, T. Foley, D. Horn, J. Sugerman, K. Fatahalian, M. Houston, and P.Hanrahan. Brook for gpus: Stream computing on graphics hardware. In ACMTransactions on Graphics (TOG) - Proceedings of ACM SIGGRAPH 2004, pages 777–786, August 2004.
[22] S. Asano, T. Maruyama, and Y. Yamaguchi. Performance comparison of fpga, gpu and cpu in image processing. In Field Programmable Logic and Applications, 2009. FPL 2009. International Conference on, pages 126 –131, 31 2009-sept. 2 2009.
[23] Shane Ryoo, Christopher I. Rodrigues, Sara S. Baghsorkhi, Sam S. Stone, David B. Kirk, and Wen-mei W. Hwu. Optimization principles and application performance evaluation of a multithreaded gpu using cuda. In Proceedings of the 13th ACM SIGPLAN Symposium on Principles and practice of parallel programming, PPoPP ’08, pages 73–82, New York, NY, USA, 2008. ACM.
[24] Amd app acceleration. http://www.amd.com/stream.
[25] Using inline ptx assembly in cuda. http://tinyurl.com/866zchu.
[26] Windows azure, microsoft’s cloud platform. http://www.windowsazure.com/.
[27] L. Shi, H. Chen, and J. Sun. vcuda: Gpu-accelerated high-performance computing in virtual machines. In IEEE International Symposium on Parallel & Distributed Processing, pages 1–11, May 2009.
[28] Xml-rpc. http://xmlrpc.com/.
[29] Extensible markup language (xml). http://www.w3pdf.com/W3cSpec/XML/2/REC-xml11-20060816.pdf.
[30] G. Giunta, R. Montella, G. Agrillo, and G. Coviello. A gpgpu transparent virtualization component for high performance computing clouds. In Euro-Par 2010 V Parallel Processing, pages 379–391, September 2010.
[31] J. Duato, A.J. Pena, F. Silla, R. Mayo, and E.S. Quintana-Orti. rcuda: Reducing the number of gpu-based accelerators in high performance clusters. In High Performance Computing and Simulation (HPCS), 2010 International Conference on, pages 224–231, August 2010.
[32] J. Duato, A.J. Pena, F. Silla, R. Mayo, and E.S. Quintana-Orti. Performance of cuda virtualized remote gpus in high performance clusters. In Parallel Processing (ICPP), 2011 International Conference on, pages 365–374, June 2011.
[33] Cuda toolkit 4.0. http://developer.nvidia.com/cuda-toolkit-40.
[34] T. Li, V.K. Narayana, E. El-Araby, and T. El-Ghazawi. Gpu resource sharing and virtualization on high performance computing systems. In Parallel Processing (ICPP), 2011 International Conference on, pages 733–742, June 2011.
[35] Kepler, next generation cuda compute architecture.
http://www.nvidia.com.tw/content/PDF/kepler/NVIDIA-Kepler-GK110-Architecture-Whitepaper.pdf.
[36] Message passing interface. http://www.mcs.anl.gov/research/projects/mpi/.