本文關(guān)鍵詞： 分級(jí)管理 存儲(chǔ)虛擬化 FV算法　出處：《國(guó)防科學(xué)技術(shù)大學(xué)》2013年碩士論文　論文類(lèi)型：學(xué)位論文

【摘要】：隨著互聯(lián)網(wǎng)和移動(dòng)通信技術(shù)的高速發(fā)展，人們對(duì)儲(chǔ)存系統(tǒng)的容量和性能的要求也越來(lái)越高，尤其隨著云計(jì)算的發(fā)展，大容量、高性能的存儲(chǔ)管理系統(tǒng)成為云計(jì)算平臺(tái)的重要組成，云存儲(chǔ)技術(shù)正在逐步走向普及，這就導(dǎo)致了高性能存儲(chǔ)設(shè)備高昂的成本與海量的數(shù)據(jù)存儲(chǔ)容量之間的矛盾也隨之日益突出；復(fù)雜的存儲(chǔ)維護(hù)及擴(kuò)容更進(jìn)一步增加了存儲(chǔ)系統(tǒng)的管理成本，分級(jí)存儲(chǔ)管理一直以來(lái)就是存儲(chǔ)管理的研究重點(diǎn)，可以有效的解決高性能與大容量之間的矛盾。 本文對(duì)分級(jí)存儲(chǔ)管理系統(tǒng)進(jìn)行研究，并最終實(shí)現(xiàn)了分級(jí)存儲(chǔ)管理系統(tǒng)(Multi-Level Files System，MLFS)原型，通過(guò)分級(jí)存儲(chǔ)和統(tǒng)一視圖虛擬化達(dá)到了合理分配系統(tǒng)資源，提高系統(tǒng)性能的目的。 首先，提出一個(gè)分級(jí)存儲(chǔ)管理模型，不僅將存儲(chǔ)設(shè)備劃分為高性能和海量存儲(chǔ)兩級(jí)，而且針對(duì)不同的分布式文件系統(tǒng)特點(diǎn)，分別使用了Lustre、GlusterFS等文件系統(tǒng)來(lái)管理，以提高性能和海量存儲(chǔ)資源，并在上述分布式文件系統(tǒng)之上，開(kāi)發(fā)了具有統(tǒng)一存儲(chǔ)資源視圖的虛擬化文件系統(tǒng)MLFS，并實(shí)現(xiàn)分級(jí)管理功能。 其次，，分析了現(xiàn)有的數(shù)據(jù)遷移算法，對(duì)數(shù)據(jù)要素考慮較為全面、數(shù)據(jù)價(jià)值判定較為準(zhǔn)確的FV算法進(jìn)行深入研究分析，并在MLFS系統(tǒng)中實(shí)現(xiàn)該算法。 最后，使用Disksim模擬器對(duì)FV算法和其他算法進(jìn)行了比較，并在搭建的真實(shí)環(huán)境中實(shí)測(cè)了MLFS系統(tǒng)的單線程和多線程I/O性能，同時(shí)對(duì)分級(jí)管理的有效性進(jìn)行了驗(yàn)證。
[Abstract]:With the rapid development of Internet and mobile communication technology, the storage system capacity and performance requirements are increasingly high, especially with the development of cloud computing, large capacity. High performance storage management system has become an important component of cloud computing platform, cloud storage technology is gradually becoming popular. This leads to the contradiction between the high cost of the high performance storage device and the massive data storage capacity. The complex storage maintenance and expansion further increase the management cost of storage system. Hierarchical storage management has been the research focus of storage management. It can effectively solve the contradiction between high performance and large capacity. This paper studies the hierarchical storage management system, and finally implements the prototype of hierarchical storage management system named Multi-Level Files system. Through hierarchical storage and unified view virtualization, the system resources are allocated reasonably and the system performance is improved. Firstly, a hierarchical storage management model is proposed, which not only divides storage devices into two levels: high performance and mass storage, but also uses Lustre according to the characteristics of different distributed file systems. File systems such as GlusterFS are managed to improve performance and mass storage resources on top of the above distributed file systems. A virtual file system MLFS with unified storage resource view is developed and hierarchical management is implemented. Secondly, the existing data migration algorithm is analyzed, and the FV algorithm, which takes the data elements into account and the data value is accurate, is studied and analyzed, and the algorithm is implemented in MLFS system. Finally, the Disksim simulator is used to compare FV algorithm with other algorithms, and the single-thread and multi-thread I / O performance of MLFS system is measured in the real environment. At the same time, the effectiveness of hierarchical management is verified.
【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】：2013
【分類(lèi)號(hào)】：TP333

【參考文獻(xiàn)】

相關(guān)期刊論文 前6條

1 胡娜;李戰(zhàn)懷;張曉;;分級(jí)存儲(chǔ)中自適應(yīng)選擇遷移策略的研究[J];計(jì)算機(jī)科學(xué);2012年03期

2 呂帥;劉光明;徐凱;劉欣;;海量信息分級(jí)存儲(chǔ)數(shù)據(jù)遷移策略研究[J];計(jì)算機(jī)工程與科學(xué);2009年S1期

3 鄭霄;陳忠平;吳志兵;;IB QoS的帶寬分配機(jī)制研究[J];計(jì)算機(jī)科學(xué);2012年04期

4 陳康;鄭緯民;;云計(jì)算:系統(tǒng)實(shí)例與研究現(xiàn)狀[J];軟件學(xué)報(bào);2009年05期

5 趙鐵柱;董守斌;Verdi MARCH;Simon SEE;;面向并行文件系統(tǒng)的性能評(píng)估及相對(duì)預(yù)測(cè)模型[J];軟件學(xué)報(bào);2011年09期

6 邊根慶;王艷云;邵必林;于泳浩;;基于價(jià)值評(píng)估的海量數(shù)據(jù)遷移策略研究[J];西安建筑科技大學(xué)學(xué)報(bào)(自然科學(xué)版);2012年03期

相關(guān)博士學(xué)位論文 前1條

1 侯f ;海量網(wǎng)絡(luò)存儲(chǔ)系統(tǒng)中的多級(jí)緩存技術(shù)研究[D];華南理工大學(xué);2011年

本文編號(hào)：1491860