The Research Of Data Security Of The Cloud Storage System Based On Hadoop Distributed File System

In the version of Hadoop0.20.203, HDFS (Hadoop Distributed File System), which employs the master-slaves structure, is consisted by one Namenode, one SecondaryNamenode and lots of Datanode. As a single master node in HDFS, Namenode has some drawbacks, such as single point of failure, performance bottlenecks, and the disadvantage of expansibility. At the same time, the main purpose of the HDFS is building a distributed file storage cluster by some cheap hosts and server. In this condition hardware failure is common.To improve the security and the efficiency of the Namenode, this paper mainly includes the following aspects of the work:1. Introduce the concept of the HDFS, including the course of development, problems existing and Research Summary Lastly;2. Introduce the details of HDFS System component, including the Namenode and Datanode. Doing the depth study of Meta data, the organization and interaction and maintain of the data block;3. Propose the Novel Distributed Namenode Cluster Program. In this distributed Namenode cluster, the function of the Namenode is reallocated. The Namenode1cluster is mainly used to cope with the client’s request and maintain the status of the Datanode; the function of the Namenode2cluster is to manage and persist the Meta data, at the same time, managing the data blocks. The Leadernode is mainly used to accept the client’s request and maintain the status of the whole Distributed Namenode Cluster, at the same time, returning the response to the client. We study the DRBD and Pacemaker. We take the shortcoming of the Namenode Program existing into seriously consideration. At the same time, the single Namenode in the HDFS is configured with HA function and testing it;4. Introduce the redundancy mechanism in the Distributed file system. We do the depth research of the redundancy mechanism in HDFS and verify the effection. Meanwhile, We study the influence on the interacting of data and load balancing by data redundancy mechanism in HDFS.5. Concluding the paper and proposing some area for improvement.