云计算下虚拟机部署机制的研究
|
摘要
云计算是指可以随时随地、按需、便捷地利用互联网访问共享资源池的一种计算模式。用户只需要支付实际资源的使用费用,就可以根据其业务向云平台申请所需的服务。虚拟化技术是应用去耦合技术将底层的硬件设备和上层的软件服务分离,以此实现重新整合和划分计算机物理资源的目标。因此,虚拟化是云计算的核心技术。
在云计算环境下,虚拟机部署系统逻辑上分为三个功能部分：核心管理中心、虚拟机镜像模板库和服务器集群。核心管理中心接受用户服务请求,在虚拟机镜像模板库中选择所需的模板,从虚拟机镜像模板库中加载至服务器集群上,最后通过将镜像模板文件实例化来创建出满足用户需求的虚拟机。
然而,随着需求服务要求的提高,特别是对于专业用户而言,虚拟机镜像模板文件非常大,通常能达到十几甚至几十个G,这带来的局限性有以下两方面：(1)这么大的文件传输要耗费大量时间,研究表明虚拟机镜像模板文件的传输时间占用虚拟机部署的大部分时间；(2)导致较高的网络开销,特别是中央存储中心的传输能力成为了整个系统性能的瓶颈。因此,对于虚拟机快速部署机制的研究成为了人们研究的热点。
本文对当前国内外关于云计算下虚拟机部署的机制展开了广泛的调研,并深入比较了一些提高虚拟机部署效率的方法,同时,研究了虚拟机镜像模板并行传输机制的优化策略。在以上的调研结果下,提出了一种优化方案。采用将虚拟机镜像模板文件分割为若干数据块,分散存储在各个宿主机上,在模板数据块文件从云存储中心加载至宿主机上的过程中,采用优化的并行传输的方法。这样带来两方面的好处：(1)减少虚拟机镜像模板文件的传输时间；(2)本方法将镜像模板文件以分布式方式存储在多个服务器上,缓解了单中心传输的压力。
本文在太原理工大学信息中心机房的服务器、存储、网络等设备上进行了实验。在服务器上安装VMware Server2.0为虚拟机监视器,此外还安装有模板文件分割模块、数据块管理模块、数据块并行传输模块和虚拟机实例化模块等核心功能模块。实验分析了在传统部署策略下和本文策略下的虚拟机部署效率。实验结果显示,随着用户部署的虚拟台数增加时,该方法在部署效率上的优越性得到体现。
Cloud computing is a computing model that can access the shared resource pool through Internet conveniently on the demand of users anywhere, anytime. Users can apply service according with their own requirements by paying the actual cost of the resource. The virtualization technology is applied decoupling techniques underlying hardware devices and software services of the upper separation, in order to achieve more efficient and flexible application of computer resources. The goal of virtualization technology is the re-integration and classification of computer and physical resources. Virtualization is the core technology of cloud computing.
In the cloud computing environment, the virtual machine deployment system is logically divided into three functional parts:the core management center, a virtual machine image template library and server clusters. Core management centers for customer service requests, select the desired template is loaded from the virtual machine image template library to the server cluster in the virtual machine image template library, and finally through to mirror the instance of the template file to create to meet user needs virtual machine.
However, with improved demand service requirements, especially for professional users, the virtual machine image template file is very large, usually can reach more than a dozen or even dozens of G, which brought about the limitations of the following two aspects:(1)to transfer such a large file spends a lot of time, studies have shown that the virtual machine image template file transmission time occupied most of the time virtual machine deployed;(2) result in higher network overhead, especially the transmission capacity of the central storage center as the entiresystem performance bottlenecks. Therefore, the improved virtual machine image template file transfer mechanism can effectively improve the efficiency of the virtual machine in the cloud computing deployment.
This paper has researched on virtual machine deployment under cloud computing, and comparison of some of the deployment of efficient methods to improve virtual machine deeply. At the same time, has researched the virtual machine image template parallel optimization strategy of the transport mechanism. In the above research results, we introduce an optimization program. The virtual machine image template file is divided into a number of data blocks, decentralized storage on each host in the process template data files from the load to the host of the cloud storage centers on the use of optimized parallel transmission method. There are two advantages:(1) reduce the transmission time of the virtual machine image template file;(2) This method will mirror the template file is stored on multiple servers in a distributed manner to ease the pressure on the transmission of a single center.
We use servers, storage, network equipment and other hardware resources in Center of Information, Taiyuan University of Technology. Install the VMware Server2.0as Virtual machine management system, Segmentation module template files, data management module, data block parallel transmission module and the virtual machine instantiated modules as the core management of the system as the experimental platform. The experimental results show that, along with the increase of the virtual number of units deployed by the user, the method to be reflected in the deployment efficiency superiority.
在云计算环境下,虚拟机部署系统逻辑上分为三个功能部分：核心管理中心、虚拟机镜像模板库和服务器集群。核心管理中心接受用户服务请求,在虚拟机镜像模板库中选择所需的模板,从虚拟机镜像模板库中加载至服务器集群上,最后通过将镜像模板文件实例化来创建出满足用户需求的虚拟机。
然而,随着需求服务要求的提高,特别是对于专业用户而言,虚拟机镜像模板文件非常大,通常能达到十几甚至几十个G,这带来的局限性有以下两方面：(1)这么大的文件传输要耗费大量时间,研究表明虚拟机镜像模板文件的传输时间占用虚拟机部署的大部分时间；(2)导致较高的网络开销,特别是中央存储中心的传输能力成为了整个系统性能的瓶颈。因此,对于虚拟机快速部署机制的研究成为了人们研究的热点。
本文对当前国内外关于云计算下虚拟机部署的机制展开了广泛的调研,并深入比较了一些提高虚拟机部署效率的方法,同时,研究了虚拟机镜像模板并行传输机制的优化策略。在以上的调研结果下,提出了一种优化方案。采用将虚拟机镜像模板文件分割为若干数据块,分散存储在各个宿主机上,在模板数据块文件从云存储中心加载至宿主机上的过程中,采用优化的并行传输的方法。这样带来两方面的好处：(1)减少虚拟机镜像模板文件的传输时间；(2)本方法将镜像模板文件以分布式方式存储在多个服务器上,缓解了单中心传输的压力。
本文在太原理工大学信息中心机房的服务器、存储、网络等设备上进行了实验。在服务器上安装VMware Server2.0为虚拟机监视器,此外还安装有模板文件分割模块、数据块管理模块、数据块并行传输模块和虚拟机实例化模块等核心功能模块。实验分析了在传统部署策略下和本文策略下的虚拟机部署效率。实验结果显示,随着用户部署的虚拟台数增加时,该方法在部署效率上的优越性得到体现。
Cloud computing is a computing model that can access the shared resource pool through Internet conveniently on the demand of users anywhere, anytime. Users can apply service according with their own requirements by paying the actual cost of the resource. The virtualization technology is applied decoupling techniques underlying hardware devices and software services of the upper separation, in order to achieve more efficient and flexible application of computer resources. The goal of virtualization technology is the re-integration and classification of computer and physical resources. Virtualization is the core technology of cloud computing.
In the cloud computing environment, the virtual machine deployment system is logically divided into three functional parts:the core management center, a virtual machine image template library and server clusters. Core management centers for customer service requests, select the desired template is loaded from the virtual machine image template library to the server cluster in the virtual machine image template library, and finally through to mirror the instance of the template file to create to meet user needs virtual machine.
However, with improved demand service requirements, especially for professional users, the virtual machine image template file is very large, usually can reach more than a dozen or even dozens of G, which brought about the limitations of the following two aspects:(1)to transfer such a large file spends a lot of time, studies have shown that the virtual machine image template file transmission time occupied most of the time virtual machine deployed;(2) result in higher network overhead, especially the transmission capacity of the central storage center as the entiresystem performance bottlenecks. Therefore, the improved virtual machine image template file transfer mechanism can effectively improve the efficiency of the virtual machine in the cloud computing deployment.
This paper has researched on virtual machine deployment under cloud computing, and comparison of some of the deployment of efficient methods to improve virtual machine deeply. At the same time, has researched the virtual machine image template parallel optimization strategy of the transport mechanism. In the above research results, we introduce an optimization program. The virtual machine image template file is divided into a number of data blocks, decentralized storage on each host in the process template data files from the load to the host of the cloud storage centers on the use of optimized parallel transmission method. There are two advantages:(1) reduce the transmission time of the virtual machine image template file;(2) This method will mirror the template file is stored on multiple servers in a distributed manner to ease the pressure on the transmission of a single center.
We use servers, storage, network equipment and other hardware resources in Center of Information, Taiyuan University of Technology. Install the VMware Server2.0as Virtual machine management system, Segmentation module template files, data management module, data block parallel transmission module and the virtual machine instantiated modules as the core management of the system as the experimental platform. The experimental results show that, along with the increase of the virtual number of units deployed by the user, the method to be reflected in the deployment efficiency superiority.
引文
[1]刘鹏.云计算[M].北京：电子工业出版社,2010：1-5,36,77,114.
[2]吕斯特.虚拟化技术指南[M].北京：机械工业出版社,2011：9-20.
[3]L.Litty, D.Lie.Computer Meteorology:Monitoring Compute Clouds www.eecg.toronto.edu/～lie/papers/litty-hotos2009-web.pdf.
[4]Christopher Clark,Keir Fraser, et al. Live Migration of Virtual Maehines[A]. In:NSDI'05 Proeeedings of the 2nd conference on SymPosium on Networked Systems Design & ImPlementation-Volume2[C]. Berkeley:NSDI,2005:269-279.
[5]Emanuel Arzuaga,David R.Kaeli,et al. Quantifying load imbalance on virtualized enterprise servers[A]. In:WOSP/SIPEW' 10[C]. New York:WOSP/SIPEW,2010:229—239.
[6]Gerori,B, Fujita,H, Ishikawa,Y.An Efficient Process Live Migration Mechanism for Load Balanced Distributed Virtual Environments[A]. In:Cluster Coputing(CLUSYER),2010 IEEE International Conference[C]. Heraklion,Crete:CLUSTER,2010:200—205.
[7]董耀祖,周正伟.基于X86架构的系统虚拟机技术与应用[J].计算机工程,2006,32(13)：69-72.
[8]杨晓伟,廖光灯等.基于Xen的X86虚拟机性能调优[J].计算机工程,2006,32(24)：249-252.
[9]罗军舟,金嘉晖,宋爱波等.云计算：体系架构与关键技术[J].通信学报,2011,(07)：3-19.
[10]王海斌.动态虚拟集群部署与管理[D].吉林大学,2011.
[11]袁金艳.多虚拟机快速部署机制的研究[D].华中科技大学,2008.
[12]万林.快速部署虚拟机的新方法[J].中国教育网络,2011,(04)：64-65.
[13]VRABLE M, MA J, CHEN J, et al. Scalability, fidelity, and containment in the potemkin virtual honeyfarm[A]. SOSP'05[C]. Brighton,United Kingdom:ACM,2005.150-159.
[14]ANDR H, LAGAR-CAVILLA S, et al. SnowFlock:virtual machine cloning as a first-class cloud primitive[J]. ACM Trans Comput Syst,2011,29(1):2-39.
[15]Nicolae, B., Bresnahan, J.,et all:Efficient Multideployment and Multisnapshotting on Clouds. HPDC 2011. San Jose, CA. June 2011. http://hal.inria.fr/docs/00/57/06/82/PDF/final-paper.pdf.
[16]郑婷婷,武延军,贺也平.云计算环境下的虚拟机快速克隆技术[J].计算机工程与应用,2011,(13)：62-66.
[17]刘正伟,文中领,张海涛.云计算和云数据管理技术[J].计算机研究与发展,2012(S1)：26-31.
[18]里斯,程桦.云计算应用架构[M].北京：电子工业出版社,2010：6-9.
[19]L.Youseff,M.Butrico,D.D.Silva.Toward a Unified Ontology of Cloud Computing. www.collabogce.org/gce08/images/7/76/Lamia Youseff.pdf.
[20]A.Lenk,M.klems,J.Nimis,S.Tai.What's Inside the Cloud? An Architectural Map of the Cloud Landscape. http://www.2.computer.org/poratal/web/csdl/doi/10.1109/CLOUD.2009.5071519.
[21]Moretti C, Bulosan J, Thain D, et al. All pairs:An abstraction for data-intensive cloud computing [C] //Proc of the 22nd IEEE Int Parallel & Distributed Processing Symp (IPDPS). Piscataway, NJ:IEEE, 2008:2-9.
[22]Huan Liu,Dan Orban.GridBatch:Cloud Computing for Large-Scale Data-Intensive Batch Applications.In Eighth IEEE International Symposium on Cluster Computing and the Grid.2008.
[23]L.Z.Wang,J.Tao,M.Kunze,D.Rattu.The Cumulus Project:Build a Scientific Cloud for a Data Center.www.cca08.org/papers/Paper29-Lizhe-Wang.pdf
[24]R. P. Goldberg. Architecture of Virtual Machines. Honeywell Information Systems, Inc. Billerica. Harvard University Cambridge:Massachusetts, July 1973:29～37.
[25]李乔,郑啸.云计算研究现状综述[J].计算机科学,2011,(04)：32-37.
[26]李亚琼,宋莹,黄永兵.一种面向虚拟化云计算平台的内存优化技术[J].计算机学报,2011,(04)：685-693.
[27]秦学东.开源虚拟化——KVM的构建[J].现代图书情报技术,2011，(11)：90-93.
[28]黄景昌.Xen虚拟化技术简述[J].科技浪潮,2007,(06)：22-25.
[29]薛海峰,卿斯汉,张焕国.XEN虚拟机分析[J].系统仿真学报,2007,(23)：5556-5558.
[30]Rajkumar Buyya,Chee Shin Yeo,Srikumar Venugopal.Marketoriented cloud computing:vision, hype, and reality for delivering it services as computing utilities[C].Dalian:Proceedings of the 10th IEEE International Conference on High Performance Computing and Communications,2008:6-12.
[31]Jong-Geun Park Jin-Mee Kim Chang-won Ami Young-Choon woo Hoon Choi. Cluster Management in a Virtualized Server Environment[A]. In:Advanced Communication Technology, 2008.ICACT2008[C].Gangwon-Do:ICACT,2008:2211.
[32]王佳隽.基于云计算环境的虚拟化资源管理研究[D].复旦大学,2011
[33]About virtual machine templates[EB/OL].http://technet.microsoft.com/en-us/library /bb740838.aspx,2011.
[34]岳冬利,刘海涛,孙傲冰IaaS公有云平台调度模型研究[J].计算机工程与设计,2011,(06):1890-1893.
[35]LFS Project HomePage [EB/OL]. http://www.linuxfromseratch.org/lfs/.
[36]陈彬,肖侬,蔡志平,等.基于优化的COW虚拟块设备的虚拟机按需部署机制[J].计算机学报,2009,(10)：1914-1924.
[2]吕斯特.虚拟化技术指南[M].北京：机械工业出版社,2011：9-20.
[3]L.Litty, D.Lie.Computer Meteorology:Monitoring Compute Clouds www.eecg.toronto.edu/～lie/papers/litty-hotos2009-web.pdf.
[4]Christopher Clark,Keir Fraser, et al. Live Migration of Virtual Maehines[A]. In:NSDI'05 Proeeedings of the 2nd conference on SymPosium on Networked Systems Design & ImPlementation-Volume2[C]. Berkeley:NSDI,2005:269-279.
[5]Emanuel Arzuaga,David R.Kaeli,et al. Quantifying load imbalance on virtualized enterprise servers[A]. In:WOSP/SIPEW' 10[C]. New York:WOSP/SIPEW,2010:229—239.
[6]Gerori,B, Fujita,H, Ishikawa,Y.An Efficient Process Live Migration Mechanism for Load Balanced Distributed Virtual Environments[A]. In:Cluster Coputing(CLUSYER),2010 IEEE International Conference[C]. Heraklion,Crete:CLUSTER,2010:200—205.
[7]董耀祖,周正伟.基于X86架构的系统虚拟机技术与应用[J].计算机工程,2006,32(13)：69-72.
[8]杨晓伟,廖光灯等.基于Xen的X86虚拟机性能调优[J].计算机工程,2006,32(24)：249-252.
[9]罗军舟,金嘉晖,宋爱波等.云计算：体系架构与关键技术[J].通信学报,2011,(07)：3-19.
[10]王海斌.动态虚拟集群部署与管理[D].吉林大学,2011.
[11]袁金艳.多虚拟机快速部署机制的研究[D].华中科技大学,2008.
[12]万林.快速部署虚拟机的新方法[J].中国教育网络,2011,(04)：64-65.
[13]VRABLE M, MA J, CHEN J, et al. Scalability, fidelity, and containment in the potemkin virtual honeyfarm[A]. SOSP'05[C]. Brighton,United Kingdom:ACM,2005.150-159.
[14]ANDR H, LAGAR-CAVILLA S, et al. SnowFlock:virtual machine cloning as a first-class cloud primitive[J]. ACM Trans Comput Syst,2011,29(1):2-39.
[15]Nicolae, B., Bresnahan, J.,et all:Efficient Multideployment and Multisnapshotting on Clouds. HPDC 2011. San Jose, CA. June 2011. http://hal.inria.fr/docs/00/57/06/82/PDF/final-paper.pdf.
[16]郑婷婷,武延军,贺也平.云计算环境下的虚拟机快速克隆技术[J].计算机工程与应用,2011,(13)：62-66.
[17]刘正伟,文中领,张海涛.云计算和云数据管理技术[J].计算机研究与发展,2012(S1)：26-31.
[18]里斯,程桦.云计算应用架构[M].北京：电子工业出版社,2010：6-9.
[19]L.Youseff,M.Butrico,D.D.Silva.Toward a Unified Ontology of Cloud Computing. www.collabogce.org/gce08/images/7/76/Lamia Youseff.pdf.
[20]A.Lenk,M.klems,J.Nimis,S.Tai.What's Inside the Cloud? An Architectural Map of the Cloud Landscape. http://www.2.computer.org/poratal/web/csdl/doi/10.1109/CLOUD.2009.5071519.
[21]Moretti C, Bulosan J, Thain D, et al. All pairs:An abstraction for data-intensive cloud computing [C] //Proc of the 22nd IEEE Int Parallel & Distributed Processing Symp (IPDPS). Piscataway, NJ:IEEE, 2008:2-9.
[22]Huan Liu,Dan Orban.GridBatch:Cloud Computing for Large-Scale Data-Intensive Batch Applications.In Eighth IEEE International Symposium on Cluster Computing and the Grid.2008.
[23]L.Z.Wang,J.Tao,M.Kunze,D.Rattu.The Cumulus Project:Build a Scientific Cloud for a Data Center.www.cca08.org/papers/Paper29-Lizhe-Wang.pdf
[24]R. P. Goldberg. Architecture of Virtual Machines. Honeywell Information Systems, Inc. Billerica. Harvard University Cambridge:Massachusetts, July 1973:29～37.
[25]李乔,郑啸.云计算研究现状综述[J].计算机科学,2011,(04)：32-37.
[26]李亚琼,宋莹,黄永兵.一种面向虚拟化云计算平台的内存优化技术[J].计算机学报,2011,(04)：685-693.
[27]秦学东.开源虚拟化——KVM的构建[J].现代图书情报技术,2011，(11)：90-93.
[28]黄景昌.Xen虚拟化技术简述[J].科技浪潮,2007,(06)：22-25.
[29]薛海峰,卿斯汉,张焕国.XEN虚拟机分析[J].系统仿真学报,2007,(23)：5556-5558.
[30]Rajkumar Buyya,Chee Shin Yeo,Srikumar Venugopal.Marketoriented cloud computing:vision, hype, and reality for delivering it services as computing utilities[C].Dalian:Proceedings of the 10th IEEE International Conference on High Performance Computing and Communications,2008:6-12.
[31]Jong-Geun Park Jin-Mee Kim Chang-won Ami Young-Choon woo Hoon Choi. Cluster Management in a Virtualized Server Environment[A]. In:Advanced Communication Technology, 2008.ICACT2008[C].Gangwon-Do:ICACT,2008:2211.
[32]王佳隽.基于云计算环境的虚拟化资源管理研究[D].复旦大学,2011
[33]About virtual machine templates[EB/OL].http://technet.microsoft.com/en-us/library /bb740838.aspx,2011.
[34]岳冬利,刘海涛,孙傲冰IaaS公有云平台调度模型研究[J].计算机工程与设计,2011,(06):1890-1893.
[35]LFS Project HomePage [EB/OL]. http://www.linuxfromseratch.org/lfs/.
[36]陈彬,肖侬,蔡志平,等.基于优化的COW虚拟块设备的虚拟机按需部署机制[J].计算机学报,2009,(10)：1914-1924.