【摘要】：近年來(lái)，云計(jì)算技術(shù)發(fā)展迅速，已經(jīng)成為成IT產(chǎn)業(yè)發(fā)展的一個(gè)主流方向。在實(shí)際運(yùn)營(yíng)中，各云計(jì)算服務(wù)提供商往往需要面對(duì)大規(guī)模的虛擬化環(huán)境，需要滿足其用戶(hù)對(duì)大規(guī)模虛擬機(jī)集群的需求。由于虛擬化和分布式存儲(chǔ)等技術(shù)是整個(gè)云計(jì)算的關(guān)鍵支撐技術(shù)，因此加速虛擬機(jī)集群的部署速度可以很好地提升系統(tǒng)性能、增強(qiáng)用戶(hù)體驗(yàn)。為了加速虛擬機(jī)集群的部署，系統(tǒng)一般采用分布式存儲(chǔ)和高速緩存等技術(shù)。然而緩存只能為其所在節(jié)點(diǎn)提供數(shù)據(jù)訪問(wèn)加速的能力，不能在系統(tǒng)整體上提升數(shù)據(jù)服務(wù)能力。 本文分析了云計(jì)算環(huán)境中虛擬機(jī)鏡像存儲(chǔ)系統(tǒng)所面臨的問(wèn)題和挑戰(zhàn)，設(shè)計(jì)并實(shí)現(xiàn)了一種系統(tǒng)層的虛擬機(jī)鏡像數(shù)據(jù)的協(xié)作式緩存機(jī)制，該機(jī)制能夠充分利用各個(gè)節(jié)點(diǎn)緩存的服務(wù)能力，使其暫存的虛擬機(jī)鏡像數(shù)據(jù)能夠?yàn)槠渌?jié)點(diǎn)提供讀取服務(wù)。協(xié)作式緩存實(shí)現(xiàn)了物理節(jié)點(diǎn)間類(lèi)P2P模式的數(shù)據(jù)傳輸，達(dá)到了擴(kuò)展存儲(chǔ)系統(tǒng)數(shù)據(jù)服務(wù)能力的目的。協(xié)作式緩存的實(shí)現(xiàn)層次位于Linux內(nèi)核態(tài)，可以支持任何的VMM。實(shí)驗(yàn)結(jié)果顯示該協(xié)作式緩存機(jī)制可以有效地減少虛擬機(jī)讀取數(shù)據(jù)的傳輸開(kāi)銷(xiāo)，提升虛擬機(jī)存儲(chǔ)系統(tǒng)的負(fù)載能力。 本文還針對(duì)傳統(tǒng)的緩存替換策略在協(xié)作式緩存場(chǎng)景下的弊端，提出了基于全局信息的數(shù)據(jù)重要性權(quán)值衡量算法，，該算法根據(jù)數(shù)據(jù)塊讀取的頻度和范圍來(lái)衡量其重要性權(quán)值。并以此算法為基礎(chǔ)設(shè)計(jì)了基于數(shù)據(jù)重要性權(quán)值的緩存替換策略。通過(guò)模擬實(shí)驗(yàn)，證明了該策略的對(duì)緩存命中率的提升。
[Abstract]:In recent years, cloud computing technology has developed rapidly, has become a mainstream direction of IT industry development. In practice, cloud computing service providers often have to face large-scale virtualization environment, and need to meet the needs of their users for large-scale virtual machine clusters. Because virtualization and distributed storage are the key technologies of cloud computing, accelerating the deployment speed of virtual machine cluster can improve the system performance and enhance the user experience. In order to speed up the deployment of virtual machine cluster, distributed storage and cache are generally used in the system. However, cache can only provide the ability of data access acceleration for the node, and it can not improve the data service ability of the system as a whole. This paper analyzes the problems and challenges faced by virtual machine mirror storage system in cloud computing environment, and designs and implements a collaborative caching mechanism of virtual machine image data in system layer. This mechanism can make full use of the service ability of each node cache and make the temporary virtual machine mirror data provide read service for other nodes. Collaborative caching implements P2P-like data transmission between physical nodes and achieves the purpose of extending storage system data service capability. The implementation hierarchy of collaborative caching is in Linux kernel state and can support any VMM. The experimental results show that the cooperative caching mechanism can effectively reduce the transmission overhead of the virtual machine reading data and enhance the load capacity of the virtual machine storage system. Aiming at the disadvantages of the traditional cache replacement strategy in the cooperative cache scenario, this paper proposes a data importance weight measurement algorithm based on global information, which measures the importance weight according to the frequency and range of data block reading. Based on this algorithm, a cache replacement strategy based on the weight of data importance is designed. The simulation results show that the strategy improves the cache hit ratio.
【學(xué)位授予單位】：國(guó)防科學(xué)技術(shù)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2012
【分類(lèi)號(hào)】：TP333

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 ;VMStore: Distributed storage system for multiple virtual machines[J];Science China(Information Sciences);2011年06期

2 陳全;鄧倩妮;;云計(jì)算及其關(guān)鍵技術(shù)[J];計(jì)算機(jī)應(yīng)用;2009年09期

本文編號(hào)：2140449