Abstract: Due to brisk growth of Data Storage in Many internet Service Companies, there is always issue of regarding unstructured data storage which is generated in Terabytes [TB] and Peta bytes [PB]. Hadoop is always deal with the large amount of data Volume. Therefore increase reliability and availability should be maintained. To gain the high availability characteristic of the Hadoop and to improve failure Recovery as early as possible or failure should be avoided. The failure of the HDFS, Name node and Master Node affects the performance of the Hadoop cluster. To overcome this problem, we proposed a system which will select new recovery namenode with less amount of time which will replicate data from namenode. In this paper, we analyze behaviour of the namenode with respect to its failure and recovery from failure.

Keywords: Cloud Computing, Fault Tolerance, Hadoop, HDFS, Recovery.