The text discusses the challenges of testing the performance of the ClickHouse database management system by generating realistic test data. The author faced difficulties in creating anonymized and compressed data that preserved the original properties, such as compression ratio, cardinality, and probability distributions. To solve this problem, they explored various approaches, including explicit probabilistic models, neural networks, and Markov models. Ultimately, they developed a tool called clickhouse-obfuscator, which uses random permutations and parametrized Markov models to generate obfuscated data that meets their requirements. The tool is easy to use and can be applied to any database dump, making it a valuable resource for performance testing and optimization of ClickHouse algorithms.