Jiangsu Rural Credit Cooperative Union's Standardized Regulatory Data Reporting Platform Construction Project

1 Project Overview

With the deepening reform of China's financial industry, regulatory requirements in the banking sector are becoming increasingly stringent. To meet these requirements, the Jiangsu Rural Credit Union encountered difficulties with its regulatory data standardized submission system:

1. The number of reporting items and data required is substantial. Per regulatory requirements, reporting items cover over 10 categories, including CBRC off-site supervision data, Regulatory Data Standardization (EAST), customer risk statistical reports, PBOC anti-money laundering data, personal credit data, enterprise credit data, standardized deposit and loan data, payment information statistics, and wealth management and fund trusts. Since each reporting item demands different data interface formats, and each reporting system involves large data volumes, complex data organization, and overlapping data sources, building a unified regulatory data submission platform has become a vital task in financial institution IT system construction.

2. Regulatory requirements cover comprehensive business areas and involve complex business relationships. The validation rules among reports are intricate, data consistency demands are high, and the bank needs multi-department collaboration for data submission.

3. Inconsistent data standards during the construction of business systems have led to complexities in regulatory data extraction and processing, placing high demands on data storage and processing.

Therefore, adopting an advanced technical architecture and mature tools to build a unified regulatory data submission platform based on big data computing technology can significantly improve data submission quality and efficiency while effectively reducing the workload.

The Jiangsu Rural Credit Union decided to adopt an advanced technical architecture and mature tools to build a unified regulatory data submission platform based on big data computing technology. By leveraging the horizontal scalability of the MPP database architecture, it achieves linear horizontal expansion of data capacity and computing power. This also meets the regulatory requirements of comprehensive business coverage, complex business relationships, intricate report validation, and scenarios involving numerous large table join operations.

26.2 Solution

The data standardized submission system is built on a GBase 8a MPP Cluster. A configurable load balancing mechanism schedules and balances the load and parallel processing across nodes. The underlying distributed architecture is transparent to upper-layer applications; users interact with a single database system without needing to manage the load of individual nodes.

Jiangsu Rural Credit Union Data Standardized Submission System Architecture

System Architecture Description

Data Generation Layer: This is the source of data, including data generated during business operations and internal management. It primarily involves data from core banking systems, credit management systems, international settlement systems, wealth management systems, and data from systems such as human resources management.

Data Exchange Layer: Serving as the hub for inter-system data exchange, it facilitates bidirectional data flow between the data generation, data integration, and data application layers. Point-to-point data transfers between systems are not permitted; all downstream data provision must go through the data exchange layer. The data exchange layer does not generate data itself but can perform some cleansing on source data. Exchange modes include batch and real-time. Batch exchange is used for non-real-time, large-volume data exchange between the generation, integration, and application layers, while real-time exchange handles small amounts of real-time or near-real-time data. Data extraction, transformation, and loading (ETL) are performed using DataStage, deployed in a cluster architecture to improve extraction efficiency and avoid single points of failure.

Data Analysis Platform: It integrates data according to defined rules and stores it in a unified manner to facilitate subsequent querying and analysis. As the core for centralized data storage and integration, the data integration layer enables data sharing and avoids redundant storage and processing across systems.

Data Application Layer: It primarily stores raw and derived data from the data integration layer, processes it according to business requirements, and provides result data or interface data for applications. It also offers interfaces for users to access data. The layer performs in-depth analysis on the integrated data and presents results through various means such as fixed reports, multidimensional analysis, and ad-hoc queries.

Data Presentation Layer: It provides unified management of institutions, users, and roles, and generates and manages regulatory reports according to regulatory requirements. The presentation layer also offers a data supplementation interface to fill in missing data from the core business systems for regulatory submission purposes.

26.3 Application Results

• Implementation Summary

The original system databases were Oracle and DB2. The new system uses a 5-node GBase 8a cluster. The vendor provided support services for migrating data to the GBase 8a cluster, optimizing load performance, batch processing performance, and query SQL, thereby tuning the GBase 8a cluster to optimal performance and providing robust support for the submission system.

• Effects and Value

High Performance: Under identical conditions, GBase 8a MPP delivers 1.1x higher performance than Hadoop.

High Compatibility: GBase 8a MPP supports standard SQL syntax and scripts, whereas Hadoop has limited support for SQL syntax, scripts, or character sets.

Easy Migration: Migration tools are provided for transitioning from Oracle and DB2 to the GBase 8a MPP cluster, simplifying the migration process for applications.