ABSTRACT

Big data storage management is one of the most challenging issues for Hadoop cluster environments, since large amount of data intensive applications frequently involve a high degree of data access locality. In traditional approaches high-performance computing consists dedicated servers that are used to data storage and data replication. Therefore to solve the problems of Disparateness among the jobs and resources a “Disparateness-Aware Scheduling algorithm” is proposed in the cluster environment. In this research work we represent K-centroids clustering in big data mechanism for Hadoop cluster. This approach is mainly focused on the energy consumption in The Hadoop cluster, which helps to increase the system reliability. The Hadoop cluster consists of resources which are categorized for minimizing the scheduling delay in the Hadoop cluster using the K-Centroids clustering algorithm. A novel provisioning mechanism is introduced along with the consideration of load, energy, and network time. By integrating these three parameters, the optimized fitness function is employed for Particle Swarm Optimization (PSO) to select the computing node. Failure may occur after completion of the successful execution in the network. To improve the fault tolerance service, the migration of the cluster is focused on the particular failure node. This can recomputed the node by PSO and the corresponding optimal node is predicted. The experimental results exhibit better scheduling length, scheduling delay, speed up, failure ratio, energy consumption than the existing systems.

Back To Abstract