-
1.
公开(公告)号:KR101341386B1
公开(公告)日:2013-12-13
申请号:KR1020120103555
申请日:2012-09-18
Applicant: 한국과학기술원
Abstract: The present invention relates to an effective block encoding and block arrangement method for supporting the fault tolerance by improving spatial efficiency and for reducing a storage space additionally required according to the storage of copied materials in a block unit distributed file system supporting the fault tolerance based on data duplication. Data nodes of the distributed file system stores the duplicated blocks and the original data blocks. A name node organizes one or more stripes including the data blocks stored in the data nodes respectively, extracts the path of the local storage in the data blocks in order to execute an encoding command transmitted to the name node, generates parity blocks by encoding the data with the source coding, stores the other parity blocks in the data node according to a rule which guarantees the availability of the system, and restores the data block, in which the fault is generated, by decoding the organization information of the stripe including the block from the name node when the fault is generated in the stored data block. [Reference numerals] (200) Name node;(210) Meta data;(270) Block information registration;(280) Data block transmission;(340,350) Block 1;(AA,CC) Block 2;(BB,DD) Block n
Abstract translation: 本发明涉及一种有效的块编码和块排列方法,用于通过提高空间效率和减少根据复制材料的存储而额外需要的存储空间来支持容错,该块单元分布式文件系统支持基于 数据复制。 分布式文件系统的数据节点存储复制块和原始数据块。 名称节点分别组织包括存储在数据节点中的数据块的一个或多个条带,提取数据块中的本地存储器的路径,以执行发送到名称节点的编码命令,通过对数据进行编码来生成奇偶校验块 利用源编码,根据保证系统可用性的规则将其他奇偶校验块存储在数据节点中,并通过对包括块的条带的组织信息进行解码来恢复产生故障的数据块 从存储的数据块中生成故障时的名称节点。 (200)名称节点;(210)元数据;(270)块信息登记;(280)数据块传输;(340,350)块1;(AA,CC)块2;(BB,DD)块 ñ
-