Abstract: It is a very difficult task to extract characteristic data from the mass data obtained by network measurement. In this paper, the characteristics of performance data are defined and the performance data are divided into different data sets according to its metric category, the way of collecting the data, and the processing stage of the data. Then a performance data space is built. By deeply analyzing the data, the complex analysis process is divided into two stages: self-completing stage and abstracting stage so as to improve the efficiency of data analysis as a whole. Finally, based on the above analysis of performance data space and data sets, a distributed data analysis architecture is designed.