Abstract: Affected by factors such as privacy protection,corporate and government data are slow to be exposed.At the same time,due to the influence of network bandwidth,it is difficult for scientific research institutions to download and use massive public data.It is rare that the existing data generation tools can concurrently meet the requirements of scien-tific research work in terms of the generation of data distribution pattern,correlation,accuracy and scalability of the system.Specific to the problem of mass data generation,this paper put forward a distributed data generation model.According to the data distribution pattern and correlative relation specified in the user’s configuration,the reservoir sampling or random sampling algorithm is used for the sampling,calculation of relative relationship and splicing of the Web data knowledge base to generate the data of which the attribute accords with the user’s configuration.Through the data generation test on the distributed computing engine Apache Spark,the generated data meets the specified data distribution and correlation requirements,and the data generation speed is linear with the data size and cluster size from the statistical point of view.It shows that the data generated by the proposed data method has high accuracy and diversity of distribution,and the proposed data generation system has good scalability

Key words: Correlation calculation, Data distribution test, Data generation, Distributed computing, Reservoir sampling