the rising number of applications serving millions of users and dealing with terabytes of data need to a faster processing paradigms. Recently, there is growing enthusiasm for the notion of big data analysis. Big data analysis becomes a very important aspect for growth productivity, reliability and quality of services (QoS). Processing of big data using a powerful machine is not efficient solution. So, companies focused on using Hadoop software for big data analysis. This is because Hadoop designed to support parallel and distributed data processing. Hadoop provides a distributed file processing system that stores and processes a large scale of data. It enables a fault tolerant by replicating data on three or more machines to avoid data loss.Hadoop is based on client server model and used single master machine called NameNode. However, Hadoop has several drawbacks affecting on its performance and reliability against big data analysis. In this paper, a new framework is proposed to improve big data analysis and overcome specified drawbacks of Hadoop. These drawbacks are replication tasks, Centralized node and nodes failure. The proposed framework is called MapReduce Agent Mobility (MRAM). MRAM is developed by using mobile agent and MapReduce paradigm under Java Agent Development Framework (JADE).