Abstract—In the past few years, along with the expansion of the data volume and the cost of computer hardware down, the demand of quickly handling huge amounts of data is driving the rising and development of distributed computing. Database ^stem based on the column storage also gradually rise along with the vigorous development of the cloud computing technology. Many organizations are trying to turn the traditional row storage database migration to the column storage database, so as to adapt to the massive growth of data. In addition to providing a distributed file ^aem and supporting the MapReduce computing framework, Hadoop also provides a scalable and structured data distributed storage system: Hadoop Database(HBase). In currently industry. The HBase column store data is one of the moa popular open source products. HBase as an open source implementation of BigTable, the users pay more attention to the performance with the popularization of its application. Even though the industry has high expectations on HBase, but it am has some disadvantages, for example: not according to the characteristics of the column data stored in columns for efficient data compression, compression mode does not support direct operation data, and so on. Therefore, improving HBase performance has important practical significance. This paper analyses the principle of HBase distributed storage system first, and then studying HBase performance testing in-depth, and outlook on the future development of distributed storage system. In general, HBase will provide satisfactory read performance as long as the cluster memory is enough. But HBase write performance will be restricted by many factors. Therefore, this paper mainly studies HBase write performance. This paper introduces HBase environment firstly, then analysis of the principle of HBase distributed storage system, focus on HBase writing process analysis. This paper did the random write tea and data write tea, analyze the tea results to find the influence factors for HBase writing performance. By modifying the HBase system configuration parameters, the HBase client and the code of the writing process to achieve the goal of improve HBase write performance. Finally, summary the application of HBase combined with the practical application.