跳到主要內容

臺灣博碩士論文加值系統

(34.226.244.254) 您好!臺灣時間:2021/08/01 04:17
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:詹芳懿
研究生(外文):Jhan, Fangyi
論文名稱:基於非統一記憶體存取架構下虛擬化系統之平行化技術
論文名稱(外文):Parallel I/O On NUMA-based Virtualization Systems
指導教授:張榮貴張榮貴引用關係
指導教授(外文):Chang, Rong-Guey
口試委員:陳鵬升黃元欣薛智文
口試委員(外文):Chen, Peng-ShengHwang, Yuan-ShinHsueh, Chih-Wen
口試日期:2012-07-23
學位類別:碩士
校院名稱:國立中正大學
系所名稱:資訊工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2012
畢業學年度:100
語文別:英文
論文頁數:44
中文關鍵詞:虛擬化系統
外文關鍵詞:Virtualization
相關次數:
  • 被引用被引用:0
  • 點閱點閱:296
  • 評分評分:
  • 下載下載:2
  • 收藏至我的研究室書目清單書目收藏:0
為了能更有效的運用系統資源和節省維護成本,因此,虛擬化系統開始被廣泛的運用。然而,當虛擬化系統存取I/O時,卻會因為系統的隔離政策與安全機制而限制住效能。它不只引起應用程式的不可預測性,更降低整體的系統效能,拖累虛擬化系統的優勢。
因此在本論文中,我們提出適合網路應用與NUMA架構所設計的N-Xen (NUMA Xen)。我們利用虛擬化系統的隔離性與NUMA架構的優勢,結合二者來改進虛擬機器與記憶體的配置。在實作N-Xen方面,我們使用了三種技術。首先,我們修改了Xen在NUMA架構下的配置。再來,我們建立一個memory allocator,並搭配nemory allocation table,分配虛擬機器所需的記憶體。最後,我們更改了Xen原先對device driver的配置,微幅的修改Block I/O路徑。
實驗的結果顯示,當執行小檔案的I/O讀取時,與Base-Xen相比,N-Xen能夠提升84\%的效能;而在執行小檔案的I/O寫入時,N-Xen能改善61\%的效能。另外,當執行大檔案的I/O存取時,N-Xen皆能夠達到接近I/O平行化的效果。並且,N-Xen也會比Base-Xen增加多達125\%的CPU使用率。實驗結果證明,針對不同類型的I/O存取,N-Xen都能有效的解決虛擬化系統對於I/O存取的瓶頸。
In order to use system resource and save maintenance costs more effectively, therefore virtualization has been implemented widely in server consolidation. However, absolute isolation and safe mechanism of I/O system will generate significant overhead in virtualization. The system not only causes unpredictable applications but decreases I/O performance, hindering the advantage of virtualization.
In this paper, we propose N-Xen aiming at a virtualization system executed web applications and based on the NUMA concept. We utilize the isolation characteristic of virtualization, combining with the parallel computing advantage of NUMA. To implement N-Xen we present three techniques. First, we modify configuration of Xen on NUMA architecture. Second, we create a memory allocator with memory allocation table, assigning memory to VMs. Third, we change configuration of physical device driver and block I/O procedure of Xen.
The experiment result shows that comparing with Xen, N-Xen increases 84\% I/O throughput when 8VMs read in small size data simultaneously and increases 61\% I/O throughput if 8VMs append in same condition. Furthermore, N-Xen achieves near parallel I/O performance when VMs access sequential I/O in large-size data concurrently. N-Xen also has better CPU utilization than Base-Xen; it can achieve maximum 125\% improvement. The result proves that N-Xen can improve effectively the bottleneck of I/O system in virtualization from different types of I/O access.

1 Introduction
2 Related Work
2.1 I/O System of Xen
2.1.1 Grant Mechanism
2.1.2 Isolated Device Driver
2.1.3 Table Update
2.1.4 Direct I/O
2.2 NUMA Architecture
2.3 Virtualization with NUMA
3 Motivation
4 New NUMA Xen
4.1 System Architecture
4.1.1 Memory Allocator
4.1.2 Memory Allocation Table
4.1.3 Memory Allocator Function
4.2 I/O Flow
5 Simulation
6 Experiment
6.1 PostMark
6.1.1 Read
6.1.2 Append
6.2 Bonnie++
6.2.1 Sequential Read
6.2.2 Sequential Write
7 Conclusion

[1] A. L. Cox A. Menon and W. Zwaenepoel. Optimizing network virtualiza-
tion in xen. In Proceedings of the USENIX Annual Technical Conference,
June 2006.
[2] O. Krieger R. Harper A. Theurer, K. Rister and S. Dobbel-
stein. Virtual scalability: Charting the performance of linux in
a virtual world. In Ottawa Linux Symposium, volume 2, 2006.
http://www.linuxsymposium.org/2006/proceedings.php.
[3] M. Chapman and G. Heiser. Vnuma: A virtual shared-memory multi-
processor. In USENIX Annual Technical Conference, 2009.
[4] L. Cherkasova and R. Gardner. Measuring cpu overhead for i/o process-
ing in the xen virtual machine monitor. In USENIX Annual Technical
Conference, April 2005.
[5] Davic Chisnall. The denitive guide to the xen hypervisor. In Prentice
Hall International, 2007.
[6] S. Muthrasanallur G. Neiger G. Regnier R. Sankaran I. Schoinas R. Uhlig
B. Vembu D. Abramson, J. Jackson and J. Wiegert. Intel virtualization
technology for directed i/o. In Intel Technology Journal 10, August 2006.
http://www.intel.com/technology/itj/2006/v10i3/.
[7] M. Jeon E. Seo D. Kim, H. Kim and J. Lee. Guest-aware prioritybased
virtual machine scheduling for highly consolidated server. In Proc. Euro-
Par, 2008.
[8] A. L. Cox D. Ongaro and S. Rixner. Scheduling i/o in virtual machine
monitors. In Proceedings of VEE, 2008.
[9] Frank Denneman. http://frankdenneman.nl/category/memory/.
[10] H. Jo H. Kim and J. Lee. Xhive: Ecient cooperative caching for virtual
machines. In IEEE Transactions on Computers, June 2010.
[11] H. Blanchard J. Xenidis J. Nakajima I. Pratt, D. Magenheimer and
A. Liguori. The ongoing evolution of xen. In Proceedings of the Ottawa
Linux Symposium (OLS), volume 2, pages 255266, Ottawa, Ontario,
Canada, July 2006.
[12] J. Kristensen J. F. Kloster and A. Mejlholm. Ecient memory sharing
in the xen virtual machine monitor. In Technical Report, Department
of Computer Science, Aalborg University, January 2006.
[13] C. Kim Y. Kwon Y. Choi J. Han, J. Ahn and J. Huh. The eect of
multi-core on hpc applications in virtualized systems. In 5th Workshop
on Virtualization and High-Performance Cloud Comput., Naples, Italy,
2010.
[14] G. Janakiraman J. R. Santos and Y. Turner. Xen network i/o perfor-
mance analysis and opportunities for improvement. HP Labs, 2007.
[15] J. Janakiraman J. R. Santos, Y. Turner and I. Pratt. Bridging the
gap between software and hardware techniques for i/o virtualization. In
Proceedings of the USENIX'08 Annual Technical Conference, Boston,
Massachusset, USA, June 2008.
[16] J. R. Santos K. K. Ram and Y. Turner. Redesigning xen's memory shar-
ing mechanism for safe and ecient i/o virtualization. In Proceedings of
the 2nd Int'l Workshop on I/O Virtualization (WIOV '10), pages 29,
March 2010.
[17] Y. Turner A. L. Cox K. K. Ram, J. R. Santos and S. Rixner. Achieving
10 gb/s using safe and transparent network interface virtualization. In
Proceedings of the 2009 ACM SIGPLAN/SIGOPS international confer-
ence on Virtual execution environments, Washington, DC, USA, March
2009.
[18] C. Lameter. Local and remote mem-
ory: Memory in a linux/numa system. 2006.
ftp://ftp.kernel.org/pub/linux/kernel/people/christoph/pmig/numamemory.pdf.
[19] S. Makineni P. Apparao and D. Newell. Characterization of network
processing overheads in xen. In IEEE International Workshop on Vir-
tualization Technology in Distributed Computing (VTDC), 2006.
[20] K. Fraser S. Hand T. Harris A. Ho R. Neugebauere I. Pratt P. Barham,
B. Dragovic and A. Wareld. Xen and the art of virtualization. In
Proceedings of the nineteenth ACM symposium on Operating systems
principles (SOSP), October 2003.
[21] D. Carr A. Menon S. Rixner A. L. Cox P. Willmann, J. Shafer and
W. Zwaenepoel. Concurrent direct network access for virtual machine
monitors. In Proceedings of the 13th International Symposium on High
Performance Computer Architecture, February 2007.
[22] S. Rixner P. Willmann and A. L. Cox. Protection strategies for direct
access to virtualized i/o devices. In Proceedings of the USENIX Annual
Technical Conference (USENIX 2008), 2008.
[23] L. Xiao Q. Li, Q. Hao and Z. Li. Vm-based architecture for network
monitoring and analysis. In ICYCS '08: Proceedings of the 2008 The 9th
International Conference for Young Computer Scientists, pages 1395
1400, Washington, DC, USA: IEEE Computer Society, 2008.
[24] P. McCormicky R. Braithwaite and W. Feng. Empirical memory-access
cost models in multicore numa architectures. In Proceedings of the In-
ternational Conference on Parallel Processing, Taipei, Taiwan, 2011.
[25] D. Rao and K. Schwan. vnuma-mgr : Managing vm memory on numa
platforms. In HiPC, Goa, India, 2010.
[26] A Shah T. Deshane, M. Ben and B. Rao. Quantitative comparison of
xen and kvm. In Proceedings of the Xen Submit, 2008.
[27] K. Schwan H. Kharche N. Tolia V. Gupta, A. Gavrilovska and V. Talwar.
Gvim: Gpu-accelerated virtual machines. In HPCVirt, 2009.
[28] VMware. Vmware vsphere: The cpu scheduler in vmware esx 4.1. 2010.
www.vmware.com/les/pdf/techpaper/VMW_vSphere41_cpu_schedule_ESX.pdf.
[29] Y. Mei S. Sivathanu Y. Koh X. Pu, L. Liu and C. Pu. Understanding
performance interference of i/o workload in virtualized cloud environ-
ments. In IEEE CLOUD, pages 5158, 2010.
[30] Z. Yu Y. Dong and G. Rose. Sr-iov networking in xen: Architecture,
design and implementation. In Proceedings of the First Workshop on
I/O Virtualization (WIOV), December 2008.

QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top