Full-Stack China-Based Data Warehouse Solution for a Major State-Owned Bank
As the digital economy deepens, data—as a new factor of production—permeates operations across every industry, becoming a new driver and resource for high-quality national economic growth. The banking sector, a quintessentially data-driven industry, must accelerate the monetization of digital value to empower business development.
To build a group data governance system characterized by 'unified data, unified architecture, and unified ecosystem,' a major state-owned bank partnered closely with GBase to create a full-stack China-based data warehouse solution. This established the 'Three Horizontals, Two Verticals, and One Line' group data governance system, providing comprehensive, agile, and refined capabilities for data sharing, analytics, mining, and service presentation of digital assets.
In this 'Three Horizontals, Two Verticals, and One Line' data governance system, the 'Three Horizontals' refer to building a three-tier architecture of 'data + analytics + presentation.' The 'Two Verticals' refer to a unified, group-wide data dictionary and a standardized, end-to-end quality control mechanism, respectively. The 'One Line' denotes a clear data redline—enforcing data accountability across the board, balancing business and management regulations, tightening data discipline, and shaping a data culture.
In 2020, the bank launched a full-scale domestic IT application innovation transformation, building an enterprise data warehouse platform entirely on domestic technologies: GBase database, domestic chips, and a domestic operating system. During the data warehouse construction, the bank adopted a 'MPP database + Hadoop' technical architecture, forming a data foundation blending data lake and warehouse approaches. This enabled intelligent data exploration and services, and shaped a shared and co-built data application mechanism, ensuring sub-second presentation of the group's digital assets on both desktop and mobile devices for an immersive user experience.
Leverage the MPP database to build the foundational thematic layer and shared summary layer of the data warehouse, solving complex relational computation issues and ensuring the stability of the data warehouse model architecture and a single source of truth for data.
Use Hadoop technology to construct the source-aligned data layer, consolidate data warehouse inputs, and store data application results, exploiting its ability to ingest, store, and process polymorphic and complex-structured data. Equipped with components for massive data processing, interactive analytics, and real-time computing and access, it boosts data application efficiency and reduces implementation and operational costs.
Build data lake-warehouse application services on cloud services, capitalizing on the cloud platform's easy scalability. Object storage on IaaS provides a unified raw data storage service for the data lake, supporting horizontal scaling and accumulating the bank's data assets. PaaS-based databases and message queues underpin seamless interaction of information flows between the lake and warehouse. SaaS delivers unified data service governance over the entire service asset portfolio.
The bank deployed GBase 8a MPP Cluster to build clusters hosting dozens of applications, including the data warehouse, risk data mart, audit system, model management platform, monitoring standardized data submission platform, and reporting system. The ultra-large-scale cluster delivers massive data storage and exceptional computing power, with the ability to scale out to over a thousand nodes to support future demands. An innovation project drawing from these practical outcomes and experiences was honored as an outstanding project of the '2022 Key Research Projects' by the Jindian Innovation Application Committee.
Going forward, the bank will further expand the use of GBase across the entire organization, gradually migrating remaining data analytics applications to the GBase 8a MPP Cluster. Together with GBase, it will advance disaster recovery construction, building asynchronous and real-time active-active clusters based on synchronization tools and virtual machine mirroring. Both parties will also explore cloud deployment options, leveraging GBase Cloud Data Warehouse (GCDW), a cloud-native data warehouse, to meet the bank's elastic data warehouse system demands.