Guangdong Unicom BSS Detail Query Transformation Project — Enhancing query analysis efficiency and reducing costs

As business continues to develop, the storage scale of Guangdong Unicom's BSS historical detail records has grown rapidly. Currently, over 10 billion records are added to Guangdong Unicom's BSS detail records each month. Due to the continuous expansion of

China Unicom Guangdong BSS Record Detail Query Transformation Project – Improving Query Analysis Efficiency and Reducing Costs

1. Project Overview

  • Background

As business continues to grow, the volume of historical BSS records at China Unicom Guangdong has expanded rapidly. The BSS system now generates more than 10 billion new records each month, and this relentless data growth imposes severe cost pressure on the historical record storage system. The existing system, built on Oracle 10g, can no longer deliver efficient query responses for such large-scale data storage, and application-facing performance needs significant improvement.

  • Business Requirements

To accelerate BSS record query analysis and reduce storage costs, China Unicom Guangdong wanted to replace the existing Oracle 10g historical record database with a columnar database dedicated to storing historical records, while providing high-performance query services to external applications.

Objectives for the BSS historical record query system:

  • Replace the existing Oracle 10g historical record database with a columnar database;

  • Store at least six months of BSS historical records (excluding the current month), and where storage allows, use high compression ratios to retain even older data beyond six months;

  • Achieve highly efficient compressed-state storage to reduce storage hardware costs;

  • Deliver high-performance query and analysis response capabilities.

2. Solution

A 10-node GBase 8a MPP Cluster database environment was deployed for this project, using X86 PC Servers for all 10 nodes. Eight nodes serve as compute nodes, while two are dedicated as data loading machines for the GBase 8a MPP Cluster, improving loading performance and high availability.

Figure 2-1: Architecture diagram of the BSS record detail query transformation project for China Unicom Guangdong

3. Results and Value

Cost-effective and efficient: Runs on low-cost X86 PC servers, substantially reducing hardware investment while delivering higher processing performance than minicomputers;

Dynamic scalability: Eliminates constraints from data volume and hardware capacity; scales on demand as business data grows;

High availability: A security-group-based backup strategy ensures that node failures do not affect normal record query services;

Performance boost: Provides high-performance query capabilities with sub-second response times.