Rosselkhozbank removed the software of the American Cloudera and transferred its data lake to Russian solutions

On December 3, 2024, it became known that Rosselkhozbank completed the import substitution of the corporate data lake, replacing the software of the American company Cloudera with the products of the Russian Arenadata Group - Arenadata Hadoop and Picodata. The project started in April 2023 and was implemented without the purchase of additional equipment.

The migration was carried out in stages with a gradual decrease in the use of imported software and the addition of resources to the target cluster without significant stops to operational processes. When switching to domestic software, inconsistencies were revealed in the algorithms of the key services Hive, Yarn and Impala.

𝓲

Freepik

The director of the department big data RSHB Alexander Saburov stressed that the new platform will reduce data processing time and tighten the requirements for efficiency, providing users with high-quality information at the right time.

The Bank took a comprehensive approach to solving the problems that arose, including attracting Arenadata consulting, studying the source code of solutions, modeling at internal and partner stands, as well as integration testing. All employees were trained to work with Arenadata Hadoop at the company's training centre.

As a result of the introduction of the new platform, the bank was able to develop financial analytics, management reporting and data quality control. The platform supports analytical calculations for regional divisions and is integrated with the RAISA artificial intelligence system, which is used by more than 300 employees.

Yulia Ilyina, Director of the Arenadata Group Department for Financial Sector and International Business, noted that the replacement of a foreign solution opens up new opportunities for the financial institution to develop and meet the high requirements of customers. After the migration in the summer of 2024, the cluster was strengthened by additional computing nodes to improve fault tolerance.