Chengde Bank Data Platform Project

Project Overview

Project Background

As data from various channels continues to grow and existing data accumulates, Chengde Bank’s legacy system can no longer meet the needs of comprehensive data analysis. To address this, Chengde Bank is building a data middle platform using a Hadoop+MPP hybrid architecture. The first phase establishes an integrated platform with capabilities for data collection, preprocessing, storage, processing, analysis, and visualization, laying a foundation for further development of the data middle platform.

Business Requirements

To overcome data silos across systems, enable comprehensive channel data analysis, and provide flexibility for future data source expansion, the first-phase requirements of the data middle platform project mainly include:

  • Data Development Capability: Build a unified management platform for data exchange, job scheduling, and data development to improve technical management, enhance data development efficiency, and accelerate data flow;

  • Data Application Capability: Build an indicator management platform to enable indicator management and query within the platform;

  • Build a Raw Data Layer: Collect and aggregate data from business systems, retain original business process data as much as possible, keep data largely consistent with source systems, with only simple integration or addition of data date identifiers;

  • Build a Data Model Layer: Based on source systems, build a unified data model in accordance with the bank’s data standards, including but not limited to a basic integration layer model and a common processing layer model, to consolidate shared data, implement data standards, and form a unified, standardized business data system;

  • Build an Indicator Aggregation Layer: Based on regulatory reports, data applications, and operational management needs, sort out common indicators for financial institutions, create a practical and forward-looking indicator system, and implement at least 500 indicators on the indicator management platform;

  • Historical Data Import: Retroactively load master data files generated since January 1, 2021 into the database.

Implementation Requirements

Use encryption technology for communication data encryption; in accordance with the “Chengde Bank Data Warehouse Technical Specifications,” automatically provide business data to the data warehouse daily, with technical standards meeting the “Chengde Bank Data Quality Requirements”; support system deployment modes such as dual-machine hot standby and active-active, e.g., if the primary host fails and cannot serve externally, the standby host must immediately and automatically take over to ensure service continuity; processing capacity can be linearly scaled by adding application nodes.

Solutions

Bank of Chengde’s data middle platform integrates data from diverse heterogeneous sources, including business platforms, channels, and CRM. Data fusion is critical to linking these platforms and enabling seamless data sharing. By building a unified data storage and management platform with the GBase 8a MPP Cluster database from General Data Technology, the bank manages data from all channels centrally. Data collection and governance are carried out in compliance with the Bank of Chengde Data Warehouse Technical Specifications and Data Quality Requirements, establishing the data foundation of the data middle platform.

The first phase of the data middle platform at Bank of Chengde consists of a production environment and a test environment. In the production environment, five servers host three management nodes and five data nodes of GBase 8a, plus two management nodes and four data nodes for HADOOP. The test environment includes three hybrid nodes of GBase 8a and three hybrid nodes of HADOOP. The GBase 8a database in the production environment collects, aggregates, and processes data from business systems. Through models such as the foundational integration layer and the common processing layer, the bank precipitates public data, enforces data standards, and delivers a unified, standardized business data system, completing the first-phase data governance.

Figure 1-1: System architecture of Bank of Chengde Data Middle Platform Phase 1

Application Results

  • Implementation

Phase I of the middle platform data auditing initiative deployed 3 management nodes and 5 data nodes running GBase 8a, along with 2 management nodes and 4 data nodes running Hadoop. In Phase II, 3 additional data nodes were expanded based on business growth.

  • Outcomes & Value

Unified Data Management: Achieved centralized integration and management of Bank of Chengde’s business data, channel data, and CRM data, laying the foundation for comprehensive business analysis;

Initial Data Governance: Through consolidation and standardized cleansing, initial data governance was accomplished, meeting the People’s Bank of China’s preliminary data governance objectives;

Flexible Scalability: Leveraging a hybrid architecture, the solution achieved full data type coverage, establishing a robust groundwork for further evolution of the data middle platform.