期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2014
卷号：3
期号：7
页码：2424-2428
出版社：Shri Pannalal Research Institute of Technolgy
摘要：Storage systems are very important and basic building blocks for cloud computing technology. High performance storage servers are the better solution for cloud storage; the implementation of storage system at very low cost is a big issue. To solve this problem, the very efficient and cost effective way of storing data is to form the cluster of commodity computers. Hadoop Distributed File System (HDFS) is freeware software whose source code is freely available for cloud and it can be deployed in low-cost hardware. Cloud Storage system using the cluster computing can be build with HDFS by making optimized replication management scheme. Hadoop maintains multiple Replicas of each data file. Hadoop is using the Triplication (three replica of each block) policy to replicate the data within the cluster by using this policy the Hadoop guarantees data availability and fault tolerance. We have observed there is some large amount of data is located within the clusters which is not being used frequently and I came across the point that maintaining the three replicas of such file is not much important. It leads to poor utilization of storage space. I have read several papers of other file system and decide to implement new way of the replication called " IRM for cloud" which will create the replica of only those file that are to be needed very frequently and keeping the replica on the requesting node may decrease access time and guarantees the full utilization of the storage space. This system provides optimal replica number as well as minimizes the access time required for the nodes to access replicas by placing the data blocks at the requesting node.
