發(fā)布時間：2018-04-13 00:02

  本文選題：分布式塊存儲系統(tǒng) + 分區(qū)策略��； 參考：《華中科技大學》2013年碩士論文

【摘要】：隨著大數據時代的來臨，數據中心要存儲或處理的數據的規(guī)模急劇攀升，因此有必要對現有數據中心進行擴容或者建造更多大規(guī)模數據中心，伴隨而來的是能耗的大幅增長。在能源緊缺的今天，能耗問題已經成為數據中心面臨的最大挑戰(zhàn)。繼性能、可靠性之后，能耗參數正成為衡量存儲系統(tǒng)優(yōu)劣的另一重要指標。分布式塊存儲系統(tǒng)已經成為構建數據中心的典型方式，因此研究分布式塊存儲系統(tǒng)節(jié)能技術，具有重要意義。 從芯片級到磁盤級，從陣列級到集群級，目前存儲系統(tǒng)已經有一些節(jié)能技術的研究成果，但往往具有一定的局限性。一般來說，節(jié)能技術主要通過冗余、多轉速、緩存、不平衡負載等手段來實現。針對多副本機制的分布式塊存儲系統(tǒng)，，基于現有的節(jié)能技術，綜合了輔助節(jié)點技術、一致性哈希技術、虛擬節(jié)點技術，并巧妙地運用了分區(qū)副本策略，徹底規(guī)避了其他節(jié)能技術通常會面對的最小全覆蓋問題�；谝陨霞夹g，對開源的“牧羊犬”系統(tǒng)進行了改進，實現了可節(jié)能的分布式塊存儲系統(tǒng)。改進后的塊存儲系統(tǒng)為全對稱無中心結構，依賴corosync管理集群節(jié)點，采用epoll機制處理事件請求。 通過對原系統(tǒng)和改進后的系統(tǒng)進行相同負載下的測試對比，證明改進后的系統(tǒng)的確可以在較小性能損耗的情況下達到較理想的節(jié)能效果。但目前系統(tǒng)還存在自動化程度較低等問題，還需要進一步改進。
[Abstract]:With the advent of big data era, the scale of the data centers to store or process has risen sharply, so it is necessary to expand the existing data centers or build more large-scale data centers, accompanied by a large increase in energy consumption.In today's energy shortage, energy consumption has become the biggest challenge faced by the data center.Following the performance and reliability, energy consumption parameters are becoming another important index to measure the quality of storage system.Distributed block storage system has become a typical way to construct data center, so it is of great significance to study the energy saving technology of distributed block storage system.From the chip level to the disk level, from the array level to the cluster level, there have been some research results of energy saving technology in the memory system, but they often have some limitations.Generally speaking, energy saving technology is mainly realized by redundancy, multi-speed, cache, unbalanced load and so on.For the distributed block storage system with multi-replica mechanism, based on the existing energy-saving technology, the auxiliary node technology, the consistent hash technology, the virtual node technology are integrated, and the partitioned replica strategy is used skillfully.It completely avoids the problem of minimum full coverage that other energy saving technologies usually face.Based on the above technology, the open source "Shepherd Dog" system is improved to realize the energy saving distributed block storage system.The improved block storage system is an all-symmetric centerless system, which relies on corosync to manage cluster nodes, and adopts epoll mechanism to handle event requests.Through the comparison of the original system and the improved system under the same load, it is proved that the improved system can achieve a better energy saving effect under the condition of small performance loss.However, there are still some problems in the system, such as low degree of automation, which need to be further improved.
【學位授予單位】：華中科技大學
【學位級別】：碩士
【學位授予年份】：2013
【分類號】：TP333

【參考文獻】

相關期刊論文 前1條

1 田磊;馮丹;岳銀亮;吳素貞;毛波;;磁盤存儲系統(tǒng)節(jié)能技術研究綜述[J];計算機科學;2010年09期

本文編號：1742007