摘要
云计算是近年来被提出的一种新型的计算模式,区别于传统的服务部署方式,弹性云计算可以使用户以相对较低的成本换取需要的IT基础设施服务。云计算以其独特的服务租用方式在IT领域掀起了一场极具开创性的变革。弹性云平台通过虚拟化技术将大量离散的基础硬件整合起来,这样不仅更易于管理硬件资源,而且能显著提高计算资源的使用效率,能够将更多的精力投入到业务逻辑上,减少对基础设施的投入成本、降低技术维护难度。由此可见,虚拟资源的部署在很大程度上决定了云计算平台所提供的服务质量。
在传统的虚拟机部署策略中,通常单一地依据服务器当前的CPU状况来选择目标宿主机,忽略了虚拟机占用的资源量,也未充分考虑服务器所能承受的负载量,导致宿主机性能与虚拟机上产生的负载匹配状况不佳,不可避免地带来由于资源得不到合理利用而引起的负载不均衡问题。
针对以上问题,本文做了深入研究,现将本文的工作总结如下：
(1)在弹性云环境下,为优化部署虚拟机,提出了一种宿主机自动选择模型,该模型具有对宿主机后续负载状况的预测功能,可以对服务器的CPU、内存、硬盘、网络带宽的性能值进行预测,掌握未来时段内服务器的性能。
(2)为保证云平台的负载趋于均衡,在宿主机自动选择过程中,本文还提出了负载均衡策略：在该模型中设立了一个管理层,由管理层的服务器负责处理用户对虚拟资源的申请,并对宿主机的选择过程实行统一调度管理,淘汰掉超载的主机,选择最适合的服务器作为目标宿主机。
(3)基于个性需求的宿主机自动选择算法是该模型的核心算法,能预估算虚拟机的资源消耗。根据请求资源应用背景的不同,用户可以为各项请求资源设置不同的性能值和权值,将用户的请求量化。与传统的部署方法相比,优化了请求资源和宿主机性能的匹配程度。
本文提出的虚拟机部署机制实现了用户自助选择、虚拟机自动部署的过程,请求的量化在一定程度上能满足用户的个性需求,非常适合应用到云计算中。通过实验验证：该部署机制确定的宿主机能充分满足用户的个性需求,能有效提高业务系统的性能,降低虚拟机迁移频率,使云平台的负载趋于均衡,有利于系统的稳定运行。
Cloud computing is a new calculation model, which is presented in recent years. It is different from traditional deployment method that it can obtain the infrastructure services at the lowest cost. Cloud computing launches a groundbreaking innovation with its unique IT service hiring mode in the field of IT.
Elastic Computing Cloud integrates of a large number of virtual resources through virtualization. It is not only easier to manage hardware resources, and can significantly improve the efficiency of the use of computing resources. It is beneficial for people to put more attention into business logic, reduce infrastructure investment, and the difficulty of maintaining infrastructure technology. Virtual resource deployment decides the. quality of services provided by cloud computing platform in the very great degree.
In the usual method of deploying virtual machines, hosts are usually selected only based on hosts'current CPU conditions. It ignores the amount of resources occupied by the virtual machines, and does not adequately consider the load that the hosts can bear. The above can lead to the match between the hosts'performance and the load on virtual machines in poor. Inevitably it brings load imbalance caused b.y the irrational use of resources.
In view of the above problems, this paper makes a deep research, the work of this thesis are summarized as follows:
(1) This paper presents a model of selecting hosts automatically which can optimize deployment of virtual machines in cloud computing. The model can predict follow-up load conditions on hosts, including the server's CPU, memory, hard disk, network bandwidth, in order to master servers'performance in future period of time.
(2) In order to ensure loads relatively balance in the cloud platform, we also put forward the strategy of load balancing in the process of hosts'automatic selection:we set up a management layer in the model. The management server is responsible for handling users'virtual resource application, and implements the unified management on the host selection process so that the model chooses the most suitable server as the target host with overload hosts eliminated.
(3) Host auto-selection algorithm based on the individual needs is the core algorithm of the model, which can pre-estimates resource consumption. The users can set different performance values and weights for the requested resource according to the application background. The model can quantify the request. Compared with the traditional method of deployment, the model optimizes the match between the resource requested and performance of hosts.
The virtual machine deployment mechanism in this paper realizes the process of user self-selection, and hosts are automatically selected in the process of deployment. Quantitative requests in the model meet users'individual needs to a certain extent, so it is very suitable to apply to the cloud. Experiments show: the target host determined by the deployment mechanism can fully meet the user's individual needs, and the system's performance can also be effectively improved. It reduces the frequency of virtual machine migration, and the host clusters'load tends to be balanced, which is beneficial for the system to run stably.
引文
[1]肖斐.虚拟化云计算中资源管理的研究与实现[D].西安：西安电子科技大学,2010.
[2]JPETER MELL,TIMOTHY GRANCE. The NIST definition of cloud computing (Draft)[EB/OL]. http://csrc.nist.gov/publications/drafts/800-145/Draft-SP-800-145_cloud-definition.pdf,2011-7.
[3]RAJKUMAR BUYYA, CHEE SHIN YEO. Cloud computing and emerging IT platforms:Vision, hype, and reality for delivering computing as the 5th utility[J]. uture Generation Computer Systems,2009, 25:599-616.
[4]http://www.vmware.com/
[5]Youhui Zhang, Yanhua Li, Weimin Zheng.Automatic software deployment using user-level virtualization for cloud-computing[J]. Future Generation Computer Systems,2011.8(12).
[6]刘鹏程.云计算中虚拟机动态迁移的研究[D].上海：复旦大学,2009.
[7]POWER智慧系统的完美虚拟化技术——云计算坚实可靠的源动力[EB/OL]. http://www-31.ibm.com/ibm/cn/cloud/pdf/power_perfect.pdf,2012-03-21.
[8]刘全中.高可用系统虚拟化管理的研究与开发[D].西安：西北工业大学,2006.
[9]吴官林.高可用性虚拟化管理中心的设计与实现[D].西安：西安电子科技大学,2011.
[10]孙世昶.虚拟化技术研究及PCI设备直接访问的实现[D].大连：大连理工大学,2006.
[11]李绍.基于Linux的虚拟化技术研究和应用[D].成都：成都理工大学,2011.
[12]张婧婧VMware虚拟化技术来架构企业数据中心[D].上海：复旦大学,2010.
[13]张佳隽.基于云计算环境的虚拟化资源管理研究[D].上海：复旦大学,2011.
[14]Ramya Raghavendra,Parthasarathy Ranganathan,Vanish Talwar,Zhikui Wang, Xiaoyun Zhu. No "power" struggles:coordinated multi-level power management for the data center[C].In the 13th international conference on Architectural support for programing languages and operating systems.New York:ASPLOS,2008:48-59.
[15]Jong-Geun Park Jin-Mee Kim Chang-Won Ami Yong-Choon Woo Hoon Choi.Cluster Management in a Virtualized Server Environment[C]. Advanced Communication Technology,2008.ICACT. Gangwon-Do:ICACT,2008:2211.
[16]Christopher Clark,Keir Fraser,Steven H,Jacob Gorm Hansen,ErieJul,Christian Limpach,Ian Pratt,Andrew Warfield.Live Migration of Virtual Machines[C].NSDI'05 Proceedings of the 2nd conference on Symposium on Networked Systems Design & Implementation-Volume2. Berkeley:NSDI,2005:273-286.
[17]YiZhao,wenlong Huang.Adaptive Distributed Load Balaneing Algorithm Based on Live Migration of Virtual Maehines in Cloud[C].INC,IMS and IDC,2009. NCM'09. Seoul:NCM,2009:170-175.
[18]Gerori,B,.Fujita,H.,Ishikawa,Y.An Effieient Proeess Live Migration Mechanism for Load Balanced Distributed Virtual Environments[C].Cluster Computing(CLUSTER),2010 IEEE International Conference.Heraklion,Crete:CLUSTER,2010:197-206.
[19]Boukerehe,A.,Ming Zhang,Hengheng xie.An adaptive dynamic load balancing for large scale distributed and virtual simulations[C].Virtual Environments,Human-Computer Interfacs and Measurements Systems,2009.VECIMS'09. Hong Kong:VECIMS,2009:96-100.
[20]Xiaoqiao Meng,Pappas,V.,Li Zhang.Improving the Scalability of data Center Networks with Traffie-aware Virtual Machine Placement[C].INFOCOM. San Diego, CA:INFOCOM,2010:1-9.
[21]GuohuiWang,Ng,T.S.E. The Impact of Virtualization on Network Performance of Amazon EC2 Data Center[C].INFOCOM. San Diego,CA:INFOCOM,2010:1-9.
[22]Xiaoying Wanga,Zhihui Dua,Yinong Chenb,Sanli Lia Virtualization-based autonomic resource management for multi-tier Web applications in shared data center[J]. Journal of Systems and Software,2008,81 (9):1591-1608.
[23]Hyser Chris,Mckee Bret,Gardner Rob,Watson Brian L.Autonomic Virtual Machine Plaeement in the Data Center[EB/OL]. http://www.hpl.hp.com/techreports/2007/HPL-2007-189.html/,2007.
[24]Rodriguez-Haro,F.,Freitag,F.,Navarro,L.Autonomous Management in Virtual-Machine-Based Resource Providers[C].Self-Adaptive and Self-Organizing Systems,2008. SASO'08. Venezia:SASO, 2008:255-264.
[25]薛静.基于虚拟化的云计算平台中安全机制研究[D].西安：西北大学,2010.
[26]戴元顺.云计算技术简述[J]Research & Development,2010,31-34.
[27]刘媛媛,高庆一,陈阳.虚拟计算环境下虚拟机资源负载均衡方法[J].计算机工程,2010,36(16)：30-32.
[28]硕珺.PC集群负载均衡调度策略研究[D].营口：中国石油大学(华东),2010.
[29]袁金艳.多虚拟机快速部署机制的研究[D].武汉：华中科技大学,2008.
[30]马雪梅.分布式系统中主机负载预测[D].长春：吉林大学,2003.
[31]唐毅.网格环境中主机负载和任务执行时间预测研究[D].重庆：重庆邮电大学,2008.
[32]徐国祥.统计预测与决策[M].上海：上海财经大学出版社,2005：96—100.
[33]陈志刚,李登,曾志文.分布式系统中一种动态负载均衡策略一相关模型及算法研究[J].小型微型计算机系统,2002,23(12)：1434.
[34]LAP-SUN CHEUNG, YU-KWOK KWOK,On load balancing approaches for distributed object computing systems[J].The Journal of Supercomputing.2004,27:149-175.
[35]刘钊,黄罡,张颖,梅宏.一种自动化的资源按需管理方法[J].北京大学学报(自然科学版),2009,45(5)：784.
[36]陈亮.集群负载均衡关键技术研究[D].长沙：中南大学,2009.
[37]陈伟.异构Web集群中基于动态反馈机制的负载均衡的研究与实现[D].重庆：重庆大学,2008.
[38]http://baike.baidu.com/view/812367.htm,2011-10-10.
