English  |  正體中文  |  简体中文  |  全文笔数/总笔数 : 58237/91808 (63%)
造访人次 : 13789032      在线人数 : 60
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU Library IR team.
搜寻范围 查询小技巧:
  • 您可在西文检索词汇前后加上"双引号",以获取较精准的检索结果
  • 若欲以作者姓名搜寻,建议至进阶搜寻限定作者字段,可获得较完整数据
  • 进阶搜寻


    jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/114792


    题名: 資源負載感知之OpenStack排程機制
    其它题名: Resource usage aware scheduling in OpenStack
    作者: 賴羿廷;Lai, Yi-Ting
    贡献者: 淡江大學電機工程學系碩士班
    衛信文
    关键词: 雲端運算;排程;OpenStack;Clouds;Scheduler
    日期: 2017
    上传时间: 2018-08-03 15:04:13 (UTC+8)
    摘要: 隨著雲端服務的需求日益成長,能快速部屬雲端環境與具備低建置成本的雲端平台因而陸續出現於市面上。其中,「OpenStack」則為近期最受到關注的一套雲端平台作業系統。其內部包括了運算模組、網通模組和儲存模組三大模組,再搭配一個可以集中管理上述三大類模組的儀表板模組,最後組合成一套OpenStack共享服務,並且以提供虛擬機器的方式,對外提供運算資源以便彈性擴充或調度。使用者可以依自己的需求,選擇佈署特定模組,換句話說,人人都能下載開放原始碼並自行打造專屬的雲端基礎設施(IaaS)環境。
    然而,在雲端環境中,虛擬機應針對其目的與性質規劃出詳細所需硬體條件,並根據實體主機之規格與資源負載情形,在符合硬體條件且不影響整體執行效能條件下,選出一台最合適之主機作虛擬機資源配置。但現行OpenStack的排程演算法並沒有辦法因應使用者所訂出之詳細規格來過濾出可用之主機。此外在OpenStack中主機資源負載僅考慮到剩餘記憶體空間,但不同目的與性質的虛擬機所會消耗的硬體資源皆不同。試想當一個實體主機之CPU資源已達滿載,又或者已無剩餘網路頻寬,但系統還是將虛擬機配置於此主機,此種情況將造成系統的不穩定進而影響使用者體驗,因此OpenStack在虛擬機配置上還有很大的改進空間,這也是本篇論文所要解決的關鍵問題。
    在本篇論文中,我們首先制定出虛擬機所需詳細硬體規格,採用OpenStack內建的Host Aggregate Filter幫助我們過濾出可用實體主機後,再考慮可用主機之CPU、記憶體、硬碟與網路頻寬四種資源的使用情況,將四種條件所得之值作正規化後,相加得到此主機最終權重,最終綜合評分最高之主機配置虛擬機。藉由我們所提出的方法,能將虛擬機配置至符合所需硬體需求之主機上,並藉由四種資源條件之權重計算,確保各實體主機資源負載平衡,不會因某資源過載而影響系統穩定性與虛擬機執行效能。
    With the growing demand for cloud services, the cloud platforms which can quickly build the cloud environment with low-cost are required in the market. Now, OpenStack is an open-source, attractive cloud platform operating system, and written mainly in Python. Using OpenStack, user can build their own environment by specific modules according to their needs, in other words, everyone can download open source code and build their own unique cloud infrastructure (IaaS) environment to provide various services.
    Though the cloud environment can be built easily, the resources allocation and management of cloud platform is a challenging issue. For example, virtual machine should be configured based on detailed hardware requirements for the purpose and nature of applications in the cloud computing environment. Moreover, the system should choose the most suitable physical host for running the specified virtual machines according to its resource usage and workload. OpenStack provides basic scheduling policy for resource allocation, but OpenStack''s scheduling algorithm does not have a method to filter out available hosts with the detailed specifications required by the user. Additionally, the scheduling policy of OpenStack only takes the remaining memory space into account and ignores that different applications have different purposes and require different capabilities of the virtual machines (VMs). Thus, there is room to improve OpenStack scheduling policy for allocating virtual machines. This is also the key issue to be discussed in this paper.
    This paper first studied the services - nova-scheduler in the OpenStack core computing suite, called Nova. The main function of nova-scheduler is to allocate hardware resources for virtual machines. In the OpenStack, the user can input the parameters according to the dashboard module; customize the virtual machine specifications to obtain the entity host resources. Second, we applied Host Aggregate Filter to find out physical host that can fulfill the requirements of VMs and then proposed weighted formulas which consider usages of various resources for selecting most feasible physical host for running VM.
    The simulation results show that the proposed method can select feasible physical host for running VM while considering the load balance among hosts and ensure that the quality of services will not be downgraded due to the wrong selection.
    显示于类别:[電機工程學系暨研究所] 學位論文

    文件中的档案:

    档案 描述 大小格式浏览次数
    index.html0KbHTML73检视/开启

    在機構典藏中所有的数据项都受到原著作权保护.

    TAIR相关文章

    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - 回馈