面向多數(shù)據(jù)中心的分布式塊存儲技術(shù)研究
發(fā)布時間:2018-08-25 07:39
【摘要】:隨著數(shù)據(jù)中心的應(yīng)用越來越廣,對多數(shù)據(jù)中心數(shù)據(jù)共享的要求也越來越迫切。然而,多中心的各存儲系統(tǒng)設(shè)計(jì)架構(gòu)不同、接口迥異,給統(tǒng)一管理、有效存儲、數(shù)據(jù)檢索和更新帶來困難;另外,多中心的存儲資源分布廣泛,數(shù)據(jù)傳輸帶寬窄、速度慢,對數(shù)據(jù)訪問及其應(yīng)用的性能造成極大影響。本文針對上述問題開展研究,具體的工作包括:針對基于廣域網(wǎng)的存儲資源管理架構(gòu)特征,提出了基于塊存儲的多層次跨數(shù)據(jù)中心存儲資源管理架構(gòu)。使用塊級存儲能有效減少邏輯存儲層次,縮短I/0請求,提升系統(tǒng)性能,多層次的存儲資源管理在完成存儲資源整合的同時能將總的管理負(fù)荷分散的各個管理層上,減小管理壓力,并降低系統(tǒng)設(shè)計(jì)實(shí)現(xiàn)難度。針對跨域的多中心數(shù)據(jù)共享特點(diǎn),提出了跨數(shù)據(jù)中心的緩存方法。該方法包括讀緩存和數(shù)據(jù)快速更新,位于客戶端的讀緩存機(jī)制能有效提升系統(tǒng)性能,存儲卷級的原子更新機(jī)制在對數(shù)據(jù)進(jìn)行快速更新的同時能保持?jǐn)?shù)據(jù)的一致性。設(shè)計(jì)實(shí)現(xiàn)了多數(shù)據(jù)中心的分布式塊存儲系統(tǒng)Federal Volume Manager(FVM),進(jìn)行了性能測試。測試結(jié)果顯示,FVM在廣域網(wǎng)的環(huán)境下有較高的數(shù)據(jù)訪問性能。
[Abstract]:With the increasing application of data centers, the requirement of data sharing in multiple data centers is becoming more and more urgent. However, the design of multi-center storage system is different, and the interface is very different, which makes it difficult for unified management, effective storage, data retrieval and update. In addition, the multi-center storage resources are widely distributed, the data transmission band is narrow, and the speed is slow. It has a great impact on the performance of data access and its applications. In this paper, the above problems are studied, including: according to the characteristics of WAN-based storage resource management architecture, a multi-level cross-data center storage resource management architecture based on block storage is proposed. Using block storage can effectively reduce the level of logical storage, shorten I / 0 requests, improve system performance. Multi-level storage resource management can complete the integration of storage resources while dispersing the total management load on the various management levels. Reduce the pressure of management and reduce the difficulty of system design and implementation. According to the characteristics of multi-center data sharing across domains, a method of multi-center caching is proposed. The method includes read cache and fast update of data. The read cache mechanism located in the client can effectively improve the system performance, and the atomic updating mechanism at the storage volume level can keep the consistency of the data while updating the data quickly. A distributed block storage system (Federal Volume Manager (FVM),) with multiple data centers is designed and implemented. The test results show that FVM has high data access performance in WAN environment.
【學(xué)位授予單位】:國防科學(xué)技術(shù)大學(xué)
【學(xué)位級別】:碩士
【學(xué)位授予年份】:2013
【分類號】:TP308;TP333
,
本文編號:2202211
[Abstract]:With the increasing application of data centers, the requirement of data sharing in multiple data centers is becoming more and more urgent. However, the design of multi-center storage system is different, and the interface is very different, which makes it difficult for unified management, effective storage, data retrieval and update. In addition, the multi-center storage resources are widely distributed, the data transmission band is narrow, and the speed is slow. It has a great impact on the performance of data access and its applications. In this paper, the above problems are studied, including: according to the characteristics of WAN-based storage resource management architecture, a multi-level cross-data center storage resource management architecture based on block storage is proposed. Using block storage can effectively reduce the level of logical storage, shorten I / 0 requests, improve system performance. Multi-level storage resource management can complete the integration of storage resources while dispersing the total management load on the various management levels. Reduce the pressure of management and reduce the difficulty of system design and implementation. According to the characteristics of multi-center data sharing across domains, a method of multi-center caching is proposed. The method includes read cache and fast update of data. The read cache mechanism located in the client can effectively improve the system performance, and the atomic updating mechanism at the storage volume level can keep the consistency of the data while updating the data quickly. A distributed block storage system (Federal Volume Manager (FVM),) with multiple data centers is designed and implemented. The test results show that FVM has high data access performance in WAN environment.
【學(xué)位授予單位】:國防科學(xué)技術(shù)大學(xué)
【學(xué)位級別】:碩士
【學(xué)位授予年份】:2013
【分類號】:TP308;TP333
,
本文編號:2202211
本文鏈接:http://sikaile.net/kejilunwen/jisuanjikexuelunwen/2202211.html
最近更新
教材專著