【摘要】：隨著云計(jì)算應(yīng)用范圍的不斷擴(kuò)展,人們對(duì)于其可用性的擔(dān)憂也與日俱增。由于云計(jì)算可用性的內(nèi)在復(fù)雜性,當(dāng)前既缺乏對(duì)其可用性的高效保障機(jī)制,也缺乏對(duì)其可用性進(jìn)行量化評(píng)估的手段和方法。而且,不同單位對(duì)云計(jì)算的使用又會(huì)有其獨(dú)特的需求,現(xiàn)實(shí)需要一套面向云計(jì)算可用性的高效保障機(jī)制和量化評(píng)估方法。在對(duì)云計(jì)算現(xiàn)有可用性保障機(jī)制研究的基礎(chǔ)上,歸納了云計(jì)算可用性的影響因素�；诓煌�(lèi)型資源失效規(guī)律的不同,提出了一種基于分類(lèi)失效的多隊(duì)列資源分配機(jī)制,并針對(duì)云計(jì)算可用性缺乏量化評(píng)估手段的問(wèn)題,給出了一種云計(jì)算可用性量化評(píng)估模型及計(jì)算方法。實(shí)驗(yàn)室仿真實(shí)驗(yàn)表明,該資源分配機(jī)制可有效提高云計(jì)算的可用性,且可用性統(tǒng)計(jì)值與理論分析值基本一致。首先對(duì)云計(jì)算可用性研究現(xiàn)狀進(jìn)行分析研究,給出了云計(jì)算及其可用性的基本概念,總結(jié)了當(dāng)前云計(jì)算可用性保障機(jī)制及評(píng)估方法的特點(diǎn),重點(diǎn)研究了其存在的缺陷和不足。接著分析了云計(jì)算服務(wù)在某單位內(nèi)部三級(jí)網(wǎng)上的實(shí)際部署,討論了其存在的問(wèn)題和其明確清晰的網(wǎng)絡(luò)結(jié)構(gòu)拓?fù)淇赡軒?lái)的優(yōu)勢(shì)。其次,結(jié)合三級(jí)網(wǎng)結(jié)構(gòu)和所部署的云計(jì)算服務(wù)的特點(diǎn),針對(duì)不同層次云計(jì)算服務(wù)所需的支持資源的不同,提出了一種基于分類(lèi)失效的多隊(duì)列資源分配機(jī)制(Multi_queuing Resource Provisioning Mechanism Based on Classified Failure Rules,MRP_CFR)。MRP_CFR機(jī)制將云計(jì)算正常提供服務(wù)所需資源概要?jiǎng)澐譃橛?jì)算資源和網(wǎng)絡(luò)交互資源兩類(lèi),然后分析其各自的失效規(guī)律。在對(duì)資源進(jìn)行分配時(shí),根據(jù)不同類(lèi)型資源的失效規(guī)律,構(gòu)建區(qū)分資源類(lèi)型并支持優(yōu)先級(jí)服務(wù)的兩級(jí)多隊(duì)列資源隊(duì)列,既可保證資源的高效利用,又考慮到三級(jí)網(wǎng)部署的部分服務(wù)的特殊性。再次,依據(jù)云計(jì)算分層服務(wù)的原理,基于自定義的服務(wù)單元SU、服務(wù)集合SS等概念,提出了一種基于分層結(jié)構(gòu)的云計(jì)算可用性量化評(píng)估模型,并給出了具體的量化評(píng)估方法。其概要的流程為:根據(jù)不同的服務(wù)部署拓?fù)浣Y(jié)構(gòu),首先完成分層的可用性定量計(jì)算,然后逐層累加,給出整體的可用性量化評(píng)估方法。最后,根據(jù)MRP_CFR機(jī)制和可用性量化評(píng)估方法,在實(shí)驗(yàn)室環(huán)境下實(shí)現(xiàn)了事件發(fā)生器、虛擬資源池、資源分配控制模塊和可用性量化計(jì)算模塊等關(guān)鍵功能模塊,并根據(jù)實(shí)際的任務(wù)到達(dá)數(shù)據(jù)和服務(wù)部署拓?fù)渫瓿煽闪擞眯远吭u(píng)估。最后結(jié)合兩種應(yīng)用場(chǎng)景,進(jìn)行模擬實(shí)驗(yàn)并根據(jù)實(shí)驗(yàn)數(shù)據(jù)分析驗(yàn)證了所提出的資源分配機(jī)制和可用性定量計(jì)算方法的有效性。
[Abstract]:With the continuous expansion of cloud computing applications, people are increasingly worried about its availability. Due to the inherent complexity of cloud computing availability, there is a lack of efficient guarantee mechanism for its availability, as well as a quantitative evaluation of its availability means and methods. Moreover, different units will have their own unique requirements for the use of cloud computing. Reality needs a set of efficient guarantee mechanism and quantitative evaluation method for the availability of cloud computing. On the basis of the research on the existing availability guarantee mechanism of cloud computing, this paper sums up the influencing factors of the availability of cloud computing. Based on the different failure laws of different types of resources, a multi-queue resource allocation mechanism based on classification failure is proposed, and the lack of quantitative evaluation means for the availability of cloud computing is solved. A quantitative evaluation model and calculation method of cloud computing availability are presented. The laboratory simulation results show that the resource allocation mechanism can effectively improve the availability of cloud computing, and the availability statistics are basically consistent with the theoretical analysis values. Firstly, the research status of cloud computing availability is analyzed and studied, the basic concepts of cloud computing and its availability are given, the characteristics of current cloud computing availability guarantee mechanism and evaluation methods are summarized, and its existing defects and shortcomings are studied in detail. Then it analyzes the actual deployment of cloud computing service on a unit internal three-level network, and discusses its existing problems and the advantages that may be brought about by its clear network structure topology. Secondly, combined with the characteristics of the three-tier network structure and the deployed cloud computing services, according to the different support resources required by different levels of cloud computing services, In this paper, a multi-queue resource allocation mechanism (Multi_queuing Resource Provisioning Mechanism Based on Classified Failure Rules,MRP_CFR) based on classification failure is proposed. MRP_CFR mechanism divides the resources needed for cloud computing to provide services into computing resources and network interaction resources. Then their respective failure laws are analyzed. In the allocation of resources, according to the failure law of different types of resources, a two-level multi-queue resource queue which distinguishes resource types and supports priority services can not only ensure the efficient utilization of resources. It also takes into account the particularity of some of the services deployed by the three-level network. Thirdly, according to the principle of cloud computing hierarchical service and the concept of custom service unit SU, service set SS, a quantitative evaluation model of cloud computing availability based on hierarchical structure is proposed, and a specific quantitative evaluation method is given. The summary process is as follows: according to different service deployment topologies, the hierarchical availability quantitative calculation is completed, and then the overall availability quantitative evaluation method is given. Finally, according to MRP_CFR mechanism and availability quantitative evaluation method, the key functional modules, such as event generator, virtual resource pool, resource allocation control module and availability quantitative calculation module, are implemented in the laboratory environment. According to the actual task arrival data and service deployment topology, the availability quantitative evaluation is completed. Finally, combined with two application scenarios, the simulation experiments are carried out and the effectiveness of the proposed resource allocation mechanism and availability quantitative calculation method is verified by the analysis of the experimental data.
【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】：2014
【分類(lèi)號(hào)】：TP393.09
，

本文編號(hào)：2482464