DEVELOPING A NOVEL APPROACH FOR IMPROVING AVAILABILITY OF JOBTRACKER IN HADOOP

Authors

  • Prajakta Mane Rajarambapu Institute of Technology, Rajaramnager, Maharashtra, India
  • Prof. Ravindra J. Mandale Rajarambapu Institute of Technology, Rajaramnager, Maharashtra, India

Keywords:

JobTracker, TaskTracker,, AvatarNode,, Hadoop

Abstract

Cluster based distributed system containssingles JobTracker and multiple TaskTracker. TaskTracker performs MapReduce operation on DataNode. In this approach, JobTracker facing single point of failure problem. The solution includes theAvatarNode and ZooKeeper. Firstly, AvatarNode uses efficient checkpoint strategy to manage temporary namespace. AvatarNode gives temporary space to store the metadata about current job and flush that data after successful job execution. Secondly, ZooKeeper which is used for session monitoring by communicating with failover manager that is residing in JobTracker andAvatarNode This approach reduces Remote Procedure Call (RPC) traffic between Backup Node and DataNodeby direct communication of JobTracker and AvatarNode.

Downloads

Published

2021-03-27

Issue

Section

Articles