It is the TaskTracker’s responsibility to monitor the health status of the active task, managing its log files, detecting failures, and reporting back to the JobTracker.
Based on the Hadoop MapReduce implementation, we developed the G-Hadoop framework for MapReduce executions in a wider area.
The goal of G-Hadoop is to enable large-scale distributed computing across multiple clusters.
To share data sets across multiple administrative domains, G-Hadoop replaces HDFS, the Hadoop’s native distributed file system, with the Gfarm file system.