Company
Date Published
Author
Alexey Milovidov
Word count
2661
Language
English
Hacker News points
None

Summary

Yandex.Metrica's data storage challenges were addressed by transitioning from MyISAM to Metrage, an implementation of LSM Tree, and later to ClickHouse, a column-oriented database. Initially, Yandex.Metrica used MySQL tables with MyISAM engine, but the system had to change its approach several times due to the increasing volume of data. The transition to Metrage improved performance, but it was limited in its ability to handle custom reports. To mitigate this limitation, an OLAPServer was developed for storing custom report wizard data. ClickHouse, a column-oriented DBMS, was later developed and opensourced by Yandex to address the need for efficient ad-hoc analytics tasks with non-aggregated data. ClickHouse can store large datasets, scale linearly, and execute high-performance queries, making it suitable for web analytics tools. The system's flexibility and extensibility have made it a reliable solution for storing data in Yandex.Metrica.