本文選題：冗余編碼 + 數據重建　； 參考：《計算機科學》2017年S1期

【摘要】：為防止硬件故障或機器宕機導致的數據丟失,冗余編碼技術被廣泛應用于分布式存儲系統中來保證數據的可靠性。然而,傳統的冗余編碼技術,如里德-所羅門碼,存在著重建數據量大的問題。副本技術在重建丟失數據時只需要讀取和傳輸丟失的數據,而冗余編碼需要讀取和傳輸更大的數據量,從而消耗更多的磁盤I/O帶寬和網絡帶寬。因此,基于冗余編碼的分布式存儲系統在重建數據時將消耗更長的時間,從而將整個系統長時間暴露在一種降級的模式下,進而增加了發(fā)生永久性數據丟失的風險。為解決這個問題,減少重建數據量的冗余編碼技術不斷被提出,然而只有這些冗余編碼與傳統的里德-所羅門碼的比較,缺少它們在存儲系統的綜合比較。系統地從減少重建數據量等幾個重要方面研究了這些減少重建數據量的冗余編碼技術,從而為實際系統中采用合適的編碼提供重要參考和依據。
[Abstract]:In order to prevent data loss caused by hardware failure or machine outage, redundant coding technology is widely used in distributed storage systems to ensure the reliability of data. However, the traditional redundant coding techniques, such as Reid-Solomon code, have the problem of large amount of reconstruction data. The replica technique only needs to read and transmit the lost data while redundant coding needs to read and transmit more data so as to consume more disk I / O bandwidth and network bandwidth. Therefore, distributed storage system based on redundant coding will consume longer time when reconstructing data, thus exposing the whole system to a degradation mode for a long time, thus increasing the risk of permanent data loss. In order to solve this problem, redundant coding techniques to reduce the amount of reconstructed data have been proposed. However, only these redundant codes are compared with the traditional Reed-Solomon codes, but they lack a comprehensive comparison in the storage system. This paper systematically studies these redundant coding techniques for reducing the amount of reconstructed data from several important aspects such as reducing the amount of reconstructed data, thus providing an important reference and basis for the adoption of appropriate codes in practical systems.
【作者單位】： 海軍工程大學電子工程學院;華中科技大學計算機科學與技術學院;
【分類號】：TP333
，

本文編號：1904162