AppleCloud云计算平台的设计与实现
发布时间:2018-06-28 12:22
本文选题:云计算 + Libvirt ; 参考:《华南理工大学》2013年硕士论文
【摘要】:云计算是一种新型的资源利用和服务提供模式,它通过虚拟化和分布式技术集中管理和维护由大量计算机构成的资源池通过网络按需的向用户提供服务,如计算、存储空间和信息服务。我们把这些资源池形象地比喻为“云”。“云”通常是一些由云计算管理系统来进行维护和管理的虚拟资源,包括计算服务器、存储服务器和网络宽带资源等。云计算管理系统通过一定的结构和管理策略对这些虚拟资源进行集中管理。云计算的基本功能是通过各种接口提供对大规模数据和计算资源的访问。 作为智能计算与云计算实验室,我们需要了解和掌握云计算核心技术实现,将云计算和实验室现有的智能算法结合起来,达到充分利用实验室系统资源,。由于现有的商用云平台大多为非开源软件,不能满足实验室教学和科研的需求,所以迫切需要研发自主知识产权的云平台。 在我们对开源云平台使用和研究的过程中,逐步了解了开源云平台OpenStack、Eucalyptus的设计理念(其设计理念同亚马逊EC2非常相似且接口兼容),并对其网络结构进行了深入的分析,给出了基于虚拟机镜像注入技术的新型IP地址分配技术,并对其单一的轮询主机分配方式进行了改进,提出了动态队列主机分配方式。鉴于分布式计算已经成为云平台的下一步研究方向,我们准备在系统中引入动态可伸缩的分布式计算的云技术实现。 本文中,我们结合开源云平台OpenStack、Eucalyptus分析了目前主流的商用云计算平台AMAEC2和Google的分布式计算平台的设计理念,并基于现有的开源虚拟化方案KVM和分布式计算平台Hadoop设计并实现了自己的AppleCloud云计算平台,实现了IaaS层虚拟化资源(包括虚拟CPU、内存、硬盘、网络)的管理功能;系统采用了流行的模块化设计方法,使得系统具有灵活的的可扩展性;系统使用了业界主流的云计算开发库Libvirt进行开发,能够兼容目前大部分主流的虚拟化平台,包括KVM、Xen、VMWare、VirtualBox等;本文还创新性的引入了基于虚拟机镜像注入技术作为新型的云计算虚拟镜像IP地址分配技术,与传统的云计算系统动态IP地址分配技术,它使得云计算系统网络结构大为精简优化,对于专用的私有云而言,由于去掉了NAT网络地址转换机制,解决了传统网络管理机制所带来的网络瓶颈问题。但是另一方面,由于注入机制本身简单直接,如何安全有效的利用需要进一步的研究和探讨,另外相比与传统的IP地址分配机制,基于虚拟机镜像注入技术的IP地址分配技术在地址分配技术与具体的AMI镜像的结构相关,需要用户具有更高的AMI镜像的制作能力。根据基于虚拟机镜像注入机制的技术,我们提出并实现了云计算平台上的可视化服务部署机制,用户可以在申请虚拟机的同时选择需要部署在上面的服务,达到即时申请即时服务的效果,不需要再去定制自己的AMI镜像,,节省了时间,简化了服务部署流程,提高了服务部署的灵活度。
[Abstract]:Cloud computing is a new model of resource utilization and service providing. It manages and maintains a resource pool made up of a large number of computers through the virtualization and distribution technology to provide services to users on the network, such as computing, storage space and information services. It is often the virtual resource that is maintained and managed by the cloud computing management system, including the computing server, the storage server and the network broadband resources. The cloud computing management system manages these virtual resources through a certain structure and management strategy. The basic function of the cloud computing is to provide the large scale through various interfaces. Access to data and computing resources.
As the intelligent computing and Cloud Computing Laboratory, we need to understand and master the core technology of cloud computing, and combine the existing intelligent algorithms in the laboratory to make full use of the resources of the laboratory system. Since the existing Shang Yongyun platform is mostly non source software, it can not meet the needs of laboratory teaching and scientific research. Cloud platform with independent intellectual property rights is urgently needed.
In the process of using and studying open source cloud platform, we gradually understand the design concept of open source cloud platform OpenStack, Eucalyptus (its design concept is very similar to Amazon EC2 and interface compatibility), and its network structure is deeply analyzed, and a new IP address allocation technology based on virtual machine image injection technology is given. With the improvement of the single polling host allocation method, the dynamic queuing host allocation method is proposed. In view of the distributed computing has become the next research direction of the cloud platform, we are ready to introduce the dynamic and scalable distributed computing cloud technology into the system.
In this paper, we combine the open source cloud platform OpenStack, Eucalyptus analysis of the current mainstream commercial cloud computing platform AMAEC2 and Google distributed computing platform design concept, and based on the existing open source virtualization scheme KVM and distributed computing platform Hadoop design and implementation of their own AppleCloud cloud computing platform, the implementation of the IaaS layer The management functions of virtualized resources (including virtual CPU, memory, hard disk, network); the system has adopted a popular modular design method to make the system flexible and extensible; the system is developed by the mainstream cloud computing library Libvirt in the industry, and can accommodate most of the current mainstream virtualization platforms, including KVM, Xen, VMW Are, VirtualBox and so on; this article also introduces the innovative virtual machine image injection technology as the new cloud computing virtual image IP address allocation technology, and the traditional cloud computing system dynamic IP address allocation technology, which makes the cloud computing system network structure greatly streamlined optimization, for the private private cloud, because of the removal of the NAT The network address conversion mechanism solves the problem of network bottleneck caused by the traditional network management mechanism, but on the other hand, because the injection mechanism itself is simple and direct, it needs further research and Discussion on how to use it safely and effectively. In addition, the IP address based on the virtual machine image injection technology is compared with the traditional IP address dispenser system. The matching technology is related to the structure of the address distribution technology and the specific AMI image. It requires the user to have the ability to make a higher AMI image. Based on the technology based on the virtual machine image injection mechanism, we propose and implement a visualization service deployment mechanism on the cloud computing platform. In the above service, the effect of instant application for instant service is achieved. It does not need to customize your own AMI image again, save time, simplify the service deployment process, and improve the flexibility of service deployment.
【学位授予单位】:华南理工大学
【学位级别】:硕士
【学位授予年份】:2013
【分类号】:TP311.52
【共引文献】
相关期刊论文 前3条
1 聂佳志;;基于云计算技术的图书馆公共文化服务效能提高方案研究[J];图书馆学研究;2014年09期
2 汤赫男;;云存储中的安全问题研究及应用[J];神州;2014年17期
3 卢朝晖;;基于Windows Azure的大规模Web应用程序设计[J];计算机技术与发展;2014年12期
本文编号:2077984
本文链接:https://www.wllwen.com/falvlunwen/zhishichanquanfa/2077984.html