AN ENERGY EFFICIENT DATA PLACEMENT USING ANONYMIZATION BASED GREEDY COVER ALGORITHM IN HADOOP

The Hadoop Distributed File System (HDFS) is a big data processing tool which is used in Hadoop applications. HDFS presents unique difficulties to the current energy-conservation methods and makes it difficult to scale-down servers. We propose an energy efficient, location-aware data placement on HDFS in this paper. HDFS produces a positive count of replicas for a data block on various nodes for avoiding loss of access to the data failures in the data center. Our method is based on a greedy cover approach which efficiently utilizes the nodes by maintaining a subset of servers which will continuously on. In the covering subset every data block having at least one replica to be stored on the given nodes. Subsequently, the remaining nodes will be turned off without influencing the availability of data. Anonymization algorithm is used to maintain location privacy. The proposed algorithm proves that the MapReduce system consumes less energy than the random data placement. MATLAB Hadoop is used for evaluating the proposed algorithm, and our algorithm achieves 13% of energy reduction under the normal workloads and up to 60% of low workloads.