Innovation with Locally-Developed Databases: GBASE Deployment at a Major State-Owned Bank

Published on 2023-09-13

The big data system of a major state-owned bank is built on the GBASE database. Looking back at its construction journey: In August 2013, the big data platform was officially launched; the main database went live in 2015; in 2018, the platform won the First Prize of the Bank Technology Development Award from the People's Bank of China (PBOC); by 2019, the big data platform was completed with a main database plus eight major data marts. In 2020, the data middle platform was established and selected for the MIIT's "2020 Big Data Industry Pilot Demonstration Projects". In 2022, the Intelligent Data Service Cloud for Rural Revitalization was successfully built and received the Second Prize of the FinTech Development Award from the PBOC. Currently, the bank is focusing on building a "next-generation big data capability system for the future" around the big data platform, strengthening the data foundation, establishing operational systems, and enabling comprehensive data-driven empowerment.

In choosing the technical route for the big data platform, the bank adopted an architecture combining analytical databases with open-source distributed data processing. For the analytical database path, the bank formed a strategic partnership with GBASE, selecting GBase 8a MPP Cluster, a distributed logical data warehouse from GBASE, and integrating it with Hadoop in a converged architecture. This built a full-stack data service technology architecture around locally-developed and open-source technologies, achieving independent innovation and secure, controllable data applications.

GBase databases have seen long-term and deep application within the bank, creating a series of key benchmark cases.

As a critical MPP infrastructure, GBase 8a MPP Cluster is primarily used for storing and processing structured model data, accounting for over 70% of the data on the big data platform. The system includes the main databases BDS/GDS and eight major data marts, forming the foundational data base for the bank's big data, and supporting AI and BI platforms in the data middle platform service layer.

Data Storage

  • The main GBase database of the big data platform handles 60PB of data with 3,000+ nodes; Hadoop handles 23PB with 1,500+ nodes.

  • Data from hundreds of systems are ingested at the head office, achieving full coverage of major business systems.

Data Processing

  • Petabyte-scale structured data

  • Hundreds of thousands of batch jobs daily

  • Most full batch processing is completed before the start of the business day

  • Real-time analysis of streaming data with millisecond latency

Data Services

  • Hundreds of AI modeling initiatives have been carried out

  • Serves 8 major business areas and application scenarios

  • Tens of thousands of batch file interfaces

  • Hundreds of millions of daily visits

Beyond the main database of the big data platform, the largest GBase application scenario is the branch data mart, which currently has 10 GBase clusters with 638 nodes, capable of storing 10PB of data. From 2021 to 2023, based on the data middle platform and branch data marts, the bank promoted the migration of branch data usage scenarios to the cloud and built a comprehensive branch Data-as-a-Service (DaaS) cloud, achieving unified management, shared data, and security.

The bank's regulatory reporting data service cloud system leverages key GBase cloud data warehouse technologies to handle batch processing for over ten reporting tasks to regulatory authorities. With the increasing volume of detailed reporting data, the manual intervention stages of regulatory reporting need to handle high-concurrency online operations and massive data analysis. Preliminary research on HTAP databases, combined with adaptation to the GBase cloud data warehouse, has enabled effective isolation and elastic scaling of storage and computing resources, improving online concurrent support capabilities.

GBase is also widely used in online query services. In systems such as the bank's data middle platform BI reporting center and flexible query center, GBase query databases take processed data to provide online data queries for upper-layer applications and reports, achieving second-level responses for queries on tens of millions of data records.

Moving forward, GBASE will deepen cooperation with the bank in four key areas: IT Application Innovation transformation, security and efficiency, cloud-native, and data-friendly architectures, jointly building a future-oriented MPP and big data ecosystem.