How Otkritie Bank used Tarantool to reduce the load on backend systems by 85%

How Otkritie Bank used Tarantool to reduce the load on backend systems by 85%

Otkritie is a too-big-to-fail Russian bank, one of the top 10 largest credit institutions in Russia. It has more than 10 thousand large corporate clients, 510 thousand small and medium-sized businesses, as well as more than 3 million private customers. Otkritie has a comprehensive portfolio of digital products.

Tasks

The bank's digital system, which includes services for internal and external customers, maintains more than 57 million accounts and processes data in all physical branches of Otkritie Bank. During the merger of several banks under the Otkritie brand, a sharp increase in the volume of transactions required an upgrade of the system's throughput capacity and higher speed of data exchange between the root systems and user services.

To ensure uninterrupted data delivery and improve the user experience, Otkritie developers addressed three challenges:

  1. Increase the system's resilience to loads. Problems with processing requests were encountered as their average daily amount was growing. It was necessary to multiply the platform's stability and data availability.
  1. It was necessary to multiply the platform's stability and data availability.
  2. Speed up data exchange between internal systems and user services. Because of the phased collection of client data, its exchange could take dozens of seconds and create a significant load on the infrastructure. This negatively affected the speed of updating data in user applications.
  3. Perform the required updates without rebuilding the bank's existing IT infrastructure. Services and resources in the bank's digital system are connected by a complex logic. In addition, some services are critical to business, making it unacceptable to stop them. To avoid overhead, it was necessary to implement the solution without rebuilding the current IT landscape.

Requirements

The developers needed to:

  • Use distributed in-memory data storage with writing to disk.
  • Implement horizontal scaling capability.
  • Reduce the lag from the master system to 3 seconds.
  • Keep the data in the cache up to date (cache warming) to speed up the display of information in the application.
  • Provide autonomous operation of the frontend system when the backend systems are not available.
  • In-memory solutions class
  • Horizontal scaling
  • Lag from the master system < 3 sec
  • Support for cache warming

Solution

To address these objectives, the bank's team compared offers and products from different vendors. The short list consisted of Redis, Tarantool, and HazelCast. The products were compared based on the following criteria: price, technical support in Russia, sharding, persistence, and performance. Redis was not a good fit due to the high cost of the corporate version, and HazelCast did not have technical support in Russia. Only the Tarantool in-memory computing platform met all the criteria.

It took only two months to implement the project and fully introduce it in all Otkritie Bank branches. We set up a server with Tarantool in two data centers and refined the middle layer services so that they could access data from Tarantool. Conducted load testing and moved on to pilot operation in a production environment.

Tarantool provided the following benefits:

  • Caching business critical systems to reduce the response time for one of the bank's frontend channels.
  • Reducing the load of reading data generated by one frontend channel from the bank's backend system by 85%.
  • Protection for the bank's backend system from being overloaded during peak loads or a sharp increase in requests during marketing campaigns by storing data with the help of sharding.
  • Combining the new platform with the existing IT infrastructure without stopping and forced upgrades—the platform is deployed on top of existing systems.
  • Access to data via GraphQL and support for capturing data from Kafka and Oracle Golden Gate.

Tarantool Data Grid application in the bank's digital system

a scrollable diagram
solutions scheme

Results

  • x20

    The platform can handle peak loads that are 20 times the average daily volume of requests

  • 2 sec

    The time to update balance information on customers' accounts was reduced to two seconds instead of dozens of seconds

  • 2 months

    The data access platform was created in two months without changing the company's IT infrastructure

  • 85%

    The reading load from the bank's backend system has decreased by 85%

More
success stories

How Gazprombank speed up internal systems 50 times using the Tarantool
GPB Logo

Tell us all about your tasks, and we’ll build a tailored Tarantool-based solution

Get
a consultation

Order
a demo

Thank you for your request

Tarantool experts will contact you shortly

Thanks
Thanks