Abstract:The Hadoop distributes file system(HDFS) which can process large amounts of data effectively through large clusters. However, HDFS is designed to handle large files and suffers performance penalty while dealing with large number of small file. An approach based on MapFile is proposed to improve storage efficiency of small files in HDFS.The main idea is to add a file type judgment module while uploading a file, and create a small file queue, put the small file serialization in a MapFile container.and establishes the index file. Experimental results show that, the storage efficiency of small files is improved contrast to Hadoop Archives(HAR files).