Evaluating data storage structures of MapReduce

MapReduce framework and its open-source implementation Hadoop, a scalable and fault-tolerant infrastructure for big data analysis on large clusters, can achieve different performance with different data storage structures. This paper evaluates the performance about three kinds of data storage structures of MapReduce, namely row-store, column-store, and RCFile. The evaluating experiments are designed to test three data storage structures in terms of data loading time, data storage space, and query execution time. The experimental results show that RCFile data storage structure can achieve better performance in most cases.