Data Hub: Next-Generation Offloading System for Banking

Client: PKO Bank Polski, Poland’s leading banking institution with over 1,100 branches nationwide

Industry: Banking

Scope: Development of an offloading system that relieves strain on the bank’s core infrastructure 

Key technologies: C++, Apache Cassandra, Python, Java 

Challenge

PKO Bank Polski used to rely on large mainframe systems designed to handle everything in one place. While these systems had centralized servers with multiple processors and substantial RAM, they couldn’t handle the high volume of user requests driven by digital banking. This surge in digital engagement exposed several vulnerabilities in their mainframe-based architecture: 

• Expensive maintenance and scaling. The legacy systems couldn’t scale efficiently to meet growing demands, yet their maintenance still required substantial resources.
• Modification risks. Any changes to the mainframe carried significant risks of system failures, making updates and improvements difficult.
• Digital channel strain. The high traffic from mobile apps and other digital channels overwhelmed the core system, leading to performance bottlenecks and delays in retrieving client operations data.

Are you facing scaling limits with legacy banking infrastructure?

Reach out to our team

Solution

PKO Bank Polski engaged Neontri to develop an offloading system to manage the increased data load resulting from the adoption of digital and mobile banking. To meet this challenge, the engineering team built the following core capabilities: 

• Replication system

Our engineers created a replication system designed to capture every change made in the original banking system and reflect it in a new platform called the Data Hub. By implementing the ETL component, we ensured the solution operated as quickly as possible, achieving near-real-time data replication. 

• Distributed data storage 

Handling this volume of data required a distributed system built specifically for the task, since a traditional database would likely face the same limitations as the original system, struggling with high availability, fault tolerance, and the ability to process numerous inputs simultaneously.

To solve this, the Neontri team built a NoSQL system based on Apache Cassandra, using the DataStax Enterprise solution for robust performance, scalability, and reliability. The initial setup was a distributed cluster of 20 nodes.

• API layer

Our team recreated the original banking system’s interface as an API. This approach allowed the offloading system to “speak the same language” as the previous version, enabling seamless integration with the bank’s existing projects.

• Sophisticated search

We implemented a sophisticated search engine with fuzzy search algorithms, enabling the bank’s clients to find relevant transactions by entering just 3-4 characters. This system employs smart autocomplete functionality, which tolerates typos, slang, and abbreviations, significantly reducing search time and improving accuracy. This feature is particularly valuable for mobile app users, where the smaller screen and keyboard can limit traditional search methods. 

• Transactions enrichment

The specialized data enrichment system enables banks to provide additional transaction information to their clients. This enhancement does not affect the speed of payment processing or require customers to provide personally identifiable information. 

Here is how it works: Whenever a customer makes a purchase, the system automatically pulls up additional details about the transaction, such as the merchant’s official name, logo, website, physical address, and contact information. It also categorizes the business and pinpoints the transaction’s GPS location.

By providing this enriched data, PKO Bank Polski has improved customer satisfaction and reduced the volume of inquiries to their support channels, as users can now easily recognize and understand their transactions.

Delivery approach

Neontri started working on Data Hub with a core team of six: three C++ developers, a technical lead, an analyst, and a project manager. This team structure was designed to provide the necessary expertise to establish the system’s foundation.

Today, Data Hub is co-developed by Neontri and PKO Bank Polski. Neontri’s core unit – two C++ developers and a project manager – handles ongoing operations and maintenance, with additional developers brought in as new projects or features require.

Results

The implementation of the Data Hub has yielded measurable improvements for PKO Bank Polski, demonstrating its crucial role in the bank’s technological infrastructure:

• Enhanced efficiency. The system now handles up to 70 million offloads daily, totaling 26 TB of offloaded data, relieving the burden on the bank’s core systems while meeting high-volume demands.
• Operational resilience. The high-availability cluster, spanning 74 nodes across three data centers with six copies of each data point, delivers 99.99% SLA availability, ensuring uninterrupted service even during potential disruptions.
• Near-real-time data replication. The Data Hub processes 10 million data retrievals daily, with an average replication time of 12.2 ms, ensuring changes in the core banking system are reflected in the offloading system as they occur.
• Smooth implementation. The Data Hub seamlessly integrates with the bank’s existing ecosystem, preserving compatibility while introducing advanced capabilities.

Ready to modernize your banking infrastructure?

Neontri enables banks to reduce pressure on core systems and scale for digital growth without compromising reliability or compliance. 

Book a consultation