Jiangxi Rural Credit Union Data Warehouse Localization Upgrade Project

1. Project Overview

• Background

As an influential banking institution in Central China, Jiangxi Rural Credit Union (JXRCU) built an ODS data warehouse platform using an IBM Netezza appliance solution a few years ago. The data warehouse has effectively supported JXRCU's needs for data analysis and mining across various business scenarios, including operational analysis, decision support, and risk control.

As competition in the banking industry intensifies and various businesses develop rapidly, the data warehouse—the core competency platform for data analysis and mining—needs further evolution and advancement.

• Challenges

The Netezza appliance is based on a closed architecture, comes with a high price tag, and incurs substantial hardware and software maintenance costs over time. In actual use, it was found that while Netezza excels in batch processing, its concurrent query performance is insufficient to meet the customer's requirement for high-concurrency query scenarios.

2. Solution

JXRCU's data platform adopts a hybrid approach combining an appliance and an MPP database cluster as the foundational architecture for the data warehouse. The architecture is illustrated below (diagram not shown).

Jiangxi Rural Credit Union Data Platform Architecture (diagram omitted)

In this project, it is recommended to retain the existing Netezza appliance as the primary platform for the ODS system. By capitalizing on the appliance's high stability and advanced technologies—such as MPP parallel processing, columnar storage, compression, and FPGA hardware acceleration—for small- and medium-scale data analytics, it provides real-time or near-real-time data staging for front-end business systems along with online data processing capabilities. It primarily serves time-sensitive online business queries and small-dataset analysis scenarios, including daily and weekly reports, real-time statistics, and queries requiring high timeliness, delivering data querying and correlation analysis for fixed users and defined business scenarios.

3. Results and Value

Dynamic and Elastic Scalability: Free from limitations of data volume and hardware capacity; scale out linearly on demand as business data grows.

Independently Controllable: Proprietary intellectual property developed in China, secure, reliable, and customizable.

Low Cost: Adopts an economical hardware solution of PC servers plus local disks, reducing system construction costs. Timely and efficient after-sales service from a local vendor further lowers operational expenses.