一种云计算下针对搜索服务的虚拟机动态迁移研究

发布时间：2018-05-18 11:37

本文选题：云计算 + 搜索服务　；参考：《吉林大学》2012年硕士论文

【摘要】：随着信息和通信技术的快速发展，计算模式经历了从最初把任务集中交付给大型处理机模式，到后来发展为基于网络的分布式任务处理模式，再到最近的按需处理的云计算模式。云计算是新一代的IT模式，在后端规模庞大、非常自动化和高可靠性的云计算中心支持下，只需要接入互联网，就能非常方便地访问各种就云的应用和信息，并免去了安装和维护等烦琐操作。同时企业和个人也能以较低的价格来使用这些由云计算中心提供的服务，或者在云中直接搭建其所需的信息服务。目前，国内外都在积极研究云计算，构建云计算相关平台及应用。很多企业都通过云计算来提供各种服务，比如谷歌的搜索服务就是通过就属于谷歌的云服务之一。搜索服务已经成为大家在工作、学习、娱乐中不可或缺的工具。通过使用搜索引擎，使得我们检索信息的能力获得了极大的提高。可以说，搜索引擎是现代的计算机技术、互联网与传统的索引理论相结合的成功典范。云计算环境下搜索服务很多时候需要将任务进行调度迁移以达到更好的应用性能和服务效果。因此本文就云计算中的搜索服务提出了一个动态迁移架构和算法。动态迁移是虚拟化技术中最受关注的技术之一，也是虚拟化独有的一个技术。动态迁移就是让一台运行中的虚拟机从一台主机迁移到另一台主机，并确保在它上面运行的应用在移动过程中能正常运行。通过动态迁移，可以保证云计算中的搜索服务不间断提供服务，使用户感觉不到迁移活动。本文在深入研究云计算相关技术、搜索引擎技术、虚拟机动态迁移以及蚁群算法等基础之上，设计了一种面向搜索服务的动态迁移的架构，并提出基于虚拟机动态迁移的任务调度算法，通过比较时间来决定是否迁移，以及决定迁移后选择向哪台目标节点迁移，利用虚拟机动态迁移技术，可以把负载过重的节点上性能较差的虚拟机迁移到合适的目标节点上，以达到负载平衡的目的。算法主要包括负载均衡策略、信息策略、触发策略、定位策略和选择策略。通过实验比较，发现本文方法进行动态迁移的机制更加适用于云计算中的搜索服务。第1章主要介绍论文的研究背景，并对国内外现有的云计算平台做了详细的调查分析。第2章综述相关技术。包括云计算的相关概念、体系结构等，对搜索引擎的技术、特点和分类，虚拟机迁移技术和几种虚拟机迁移方法，以及蚁群算法等进行详细的介绍。第3章提出一种动态迁移的架构，并简要介绍动态迁移的策略。第4章优化作业调度。通过静态的方法调整优化架构性能，然后对现有作业调度算法的分析和实验，，并在此基础之上提出了两级队列期限驱动作业调度算法，以提高集群的性能。第5章介绍本文的实验环境、过程和结果的分析，验证算法的正确性。在本文最后总结本文的工作，并指出将来的研究方向。
[Abstract]:With the rapid development of information and communication technology , the computing model has experienced the most recent cloud computing model based on the network based distributed task processing mode from the initial task set to the large - scale processor mode .

Search service has become an indispensable tool in the work , study and entertainment . By using the search engine , the ability to retrieve information is greatly improved . It can be said that the search engine is a successful example of the modern computer technology , the Internet and the traditional index theory .

The search service in cloud computing environment needs to schedule and migrate tasks to achieve better application performance and service effect . Therefore , this paper presents a dynamic migration architecture and algorithm for the search service in cloud computing .

Dynamic migration is one of the most interesting technologies in virtualization and is a unique technology for virtualization . Dynamic migration is the migration of one running virtual machine from one host to another , and to ensure that the applications running on it are functioning properly during the movement . By dynamic migration , it is possible to ensure that the search service in the cloud computing is uninterrupted and that the user does not feel the migration activity .

Based on the research on cloud computing related technologies , search engine technology , virtual machine dynamic migration and ant colony algorithm , a task scheduling algorithm based on dynamic migration of virtual machines is designed , and a task scheduling algorithm based on dynamic migration of virtual machines is proposed . The virtual machines with poor performance can be migrated to the appropriate target nodes by comparing the time . The algorithm mainly includes load balancing strategy , information strategy , trigger strategy , location strategy and selection strategy . Through the experiment comparison , it is found that the mechanism of dynamic migration is more suitable for the search service in cloud computing .

Chapter 1 mainly introduces the research background of the thesis , and makes a detailed investigation and analysis on the existing cloud computing platform at home and abroad .

Chapter 2 reviews the related technologies , including the concept and architecture of cloud computing , the technology , characteristics and classification of search engines , virtual machine migration technology and several virtual machine migration methods , and ant colony algorithm .

Chapter 3 provides a framework for dynamic migration , and briefly introduces the dynamic migration strategy .

In chapter 4 , the optimal operation scheduling is optimized . The performance of the optimized architecture is adjusted by the static method , then the analysis and experiment of the existing job scheduling algorithm are carried out , and the scheduling algorithm of the two - stage queue duration driving job is put forward , so as to improve the performance of the cluster .

Chapter 5 introduces the experimental environment , process and result analysis , and verifies the correctness of the algorithm .

In the end of this article , we summarize the work of this paper and point out the future research direction .
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP302

【相似文献】