Yunnan Mobile Business Analysis System——Open, Stable, Dynamically Scalable

Project Background

In recent years, with the rapid growth of China Mobile Yunnan's business volume, the enterprise's business analysis system has faced increasing pressure from growing data processing scales. Driven by the concept of refined and personalized services, operator enterprises have developed increasingly complex processing models for business analysis, raising higher requirements for the system's data carrying capacity and analytical processing capabilities. To meet these requirements, the China Mobile Yunnan business analysis big data platform renovation project plans to introduce an MPP relational database to achieve overall performance improvements.

Solution

The MPP data warehouse database for business analysis and the traditional main database for business analysis constructed with DB are combined to form a unified data analytics platform for various subject areas, including customer analysis, service analysis, and channel analysis, for China Mobile Yunnan's business analysis operations. This lays the platform foundation for the enterprise to handle larger-scale data and more complex mining and analysis processing. As part of the expansion approach for the project's business analysis data warehouse, the specific data processing flow among the platform components after expansion is as follows:

• Collect raw business data from BOSS, billing, CRM, and other business systems;

• Clean and transform data using traditional ETL tools such as Informatica and SAS;

• Route the cleaned data via the ETL loading function to the current Oracle/DB2 business analysis main database and the MPP interface server.

The interface server completes the final data loading into the GBase 8a MPP Cluster database.

The data platform is divided into an integrated data warehouse platform built jointly by traditional centralized databases and MPP distributed databases. The new data platform can leverage the scalability advantages of MPP distributed databases while integrating the reliability guarantees of traditional data warehouse platforms, such as strong consistency for high-value density data.

Architecture Diagram of China Mobile Yunnan's Business Analysis System

Application Results

Low Cost: The GBase 8a MPP Cluster database runs on low-cost x86 PC servers, effectively reducing hardware investment and cutting expansion costs to about one-tenth of the original.

Elastic Scalability: The horizontal scaling model of GBase 8a MPP Cluster nodes allows dynamic expansion without service interruption, ensuring service continuity.

High Availability: The GBase 8a MPP Cluster database improves overall system collaboration efficiency. With a security group-based backup strategy, it ensures that the failure of any single node does not affect the continuity of services provided to the outside.