. In this work, we have presented the design and implementation of G-Hadoop, a MapReduce framework based on Hadoop that aims to enable large-scale distributed computing across multiple clusters. The architecture of G-Hadoop is based on a master/slave communication model. In order to support globally distributed data-intensive computation among multiple administrative domains, we use the traditional HDFS file system with the Gfarm file system, which can manage huge data sets across distributed clusters