Banco Tai Fung Data Platform Independently Developed Database Replacement Project

As Tai Fung Bank’s business continued to grow, data accumulation surged, yet vast amounts of historical data remained underutilized, leaving financial operations and service innovation without essential IT support. Monetizing these massive data resources to drive business value became crucial for strengthening core competitiveness and seizing market opportunities. As data volumes expanded and business scenarios evolved, the scope of data extraction and computation widened, and ETL scheduling grew increasingly complex. This not only increased operational costs but also made it difficult to ensure data quality.

Tai Fung Bank’s data platform ran on DB2 with stored data reaching the terabyte scale, facing challenges in both database performance and stability. It urgently needed to guarantee fast query responsiveness and system reliability under high concurrency and heavy loads. For scalability, the platform had to meet future business expansion requirements. From a cost perspective, it also needed to deliver high cost-effectiveness to avoid unsustainable maintenance and scaling due to excessive future spending.

To meet the data query and analysis demands of Tai Fung Bank, GBase 8a MPP Cluster (a massively parallel processing distributed database) was deployed to replace DB2 and build a new data platform, fulfilling expectations and delivering higher value.

The data platform uses a multi-node GBase 8a cluster to support terabytes of data. Through GBase 8a’s columnar storage, intelligent indexing, and high compression, disk I/O is greatly reduced, providing a significant boost in query and statistical analysis performance over the legacy system. It supports ad-hoc queries, multidimensional analysis, and various complex BI analytics and visualizations. As the foundation for the data orchestration platform, it enables unified management of models, metrics, and ETL, delivering a replicable solution for other financial industry users.

Tai Fung Bank Data Platform Architecture Diagram

Tai Fung Bank's previous data platform was DB2. It has now been fully migrated to a GBase 8a cluster, significantly improving the data warehouse report generation performance.

Massive Storage and High-Speed Loading: Achieves cluster loading of TB-level data with performance up to tens of TB per hour. High compression ratios for data ingestion further boost performance.

Ad-Hoc Queries: Enables fast, efficient ad-hoc queries with sub-second responses over large datasets, providing stable support for the scheduling system.

Statistical Analytics: Delivers significantly faster analytics performance, achieving a 2x–10x improvement over the legacy system. Complex statistical analytics respond in seconds.

Unified Management: Serves as the foundational platform for the data scheduling system, enabling unified management of models, metrics, and ETL processes.

Data Backup: Provides periodic export and backup of critical data warehouse assets, including instance-level, database-level, and table-level backups.