摘要
由于机器硬件故障导致的重要文件丢失对工作造成了极大影响。HDFS分布式文件系统通过文件副本机制提高了存储的可靠性。但HDFS中Name Node单点故障问题导致了当Name Node岩机时,整个集群不可用,从而导致了用HDFS进行文件存储不是100%的安全可靠。文中通过UCarp虚拟IP软件建立Name Node的备份节点。当HDFS岩机时,备份节点会自动竞争为新的Name Node,为集群提供元数据服务,从而保证HDFS在任何时候都不会有数据丢失现象,提高了HDFS的可靠性。
Hardware failure caused important file loss has affected our work. HDFS（ Hadoop Distributed File System） can improve the storage reliability through multi-copy mechanism. But the single point failure problem of Name Node makes the HDFS not 100% safe because when Name Node fail,the whole cluster become not available. The paper makes use of the Backup Node of Hadoop to be the backup of Name Node by the help of Ucarp virtual IP software. When Name Node fails,Backup Node will compete to be the new Name Node which provides meta data service for the whole cluster,guarantees HDFS will not loss data at any time,improving the reliability of HDFS.
出处
《西安科技大学学报》
CAS
北大核心
2016年第3期428-433,共6页
Journal of Xi’an University of Science and Technology
基金
西安科技大学2014年度教育教学改革与研究项目(JG14014)
