Jinshang Bank Independently Developed Database Procurement Project

By adopting an independently developed analytical database, Jinshang Bank leverages the query and storage capabilities of MPP database clusters. While preserving the existing data mart functions, it develops new data marts based on Jinshang’s business c

Jinshang Bank China-Developed Database Procurement Project

Project Background

Jinshang Bank adopted a China-developed analytical database, leveraging the query and storage capabilities of an MPP database cluster. Building upon the existing data mart functionalities, it developed new data mart applications tailored to the bank's business characteristics and achieved unified data integration, sharing, and presentation.

Solution

Built a data mart system for Jinshang Bank based on GBase 8a MPP Cluster. Source systems aggregate data via ETL into an analytics mart for mining, including all business reports and query business information, extracting key information.

The data cluster system comprises several components:

Common Processing Layer:

Organizes and integrates common indicators and dimensions for the data mart, standardizes data, and unifies calibers, dimensions, and indicators to provide a standard foundation for subsequent data integration.

Foundation Integration Layer:

Performs basic integration of all analytical indicators and dimensions, calculates and integrates target results, handles complex computations, stores result data, and constructs a star schema model.

Data Query Support:

Leverages the advantages of massively parallel query and columnar storage to perform data query and analysis, meeting requirements for complex associative and statistical queries while ensuring system query response times.

Results

  • Consolidated disparate systems, bridged data silos across the entire bank, and centrally managed previously fragmented resources, enabling their full utilization.

  • Unified analysis and processing, enhanced the value of customer data, and significantly reduced computing and storage costs.

  • Query response speed increased by 2 to 10 times compared to the original system, with complex queries achieving sub-second response.

  • Effectively improved real-time capabilities, with daily batch processing time reduced to one-tenth of the previous duration.

  • Easy scalability, supporting horizontal scaling on x86 infrastructure, greatly replacing the traditional vertical scaling model of standalone databases.

  • High cost-effectiveness: through horizontal scaling, total system construction costs were reduced by 50%.