China Mobile Headquarters Centralized Business Analysis System – Deep Analysis Cloud Empowers China Mobile's Business Decision-Making

Self-Service Analytics: Enable sub-second response for self-service reporting and multi-dimensional analysis queries, resolving inefficiencies in querying large datasets and significantly enhancing user satisfaction.

Elastic Scalability: The system is highly scalable, supporting dynamic cluster expansion with near-linear performance gains as nodes are added.

Cost-Efficiency: GBase 8a MPP Cluster runs on low-cost x86 PC servers, delivering high performance at minimal cost.

In-House Innovation: 100% self-developed and fully controllable, aligned with the de-IOE trend of reducing dependency on legacy proprietary infrastructure.

Hybrid Architecture for Massive Data: A hybrid architecture combining Hadoop, MPP, and a primary data warehouse, powered by distributed computing and storage, to effectively handle massive data volumes.

The Centralized Business Analytics System is built on three databases—a primary data warehouse, a GBase 8a MPP data warehouse, and a Hadoop cloud. Each database runs on an independent cluster, and data is scheduled and transferred between them via an ETL platform.

The Deep Analytics Cloud is powered by the GBase 8a MPP Cluster distributed database, with a total of 1,186 nodes (266 data warehouse nodes in Phase 1 and 920 nodes in Phase 2). It holds 13.3 PB of loaded data with a daily increase of 9.7 TB, sourced from business data uploaded across 31 provinces in China. As the core data storage platform, it consists of these four data types:

Fused data from business (B), operations (O), and management (M) domains;

Unstructured data, i.e., internet analysis and aggregated results from the Hadoop cloud;

Sandbox data and sample validation data from self-service analytics;

Data from data marts.

The data volume reaches “12+1” months of granular data.

The Deep Analytics Cloud built on the MPP database cluster provides the following key capabilities:

Lightweight aggregation of underlying structured and unstructured data, with support for multidimensional analysis, trend analysis, Top-N analysis, cause-and-effect analysis, and What-If analysis on the aggregated results;

Deep analysis and mining through associative computation across massive cross-domain datasets to generate accurate user profiles;

Self-service analysis and query capabilities;

Support for data mart applications;

Leverages the OLAP processing power of the MPP data warehouse to create data sandboxes.

Deep Analytics Cloud is positioned for historical data storage and in-depth analysis within the centralized business analytics system. While ensuring data consistency, it underpins self-service analysis and deep mining services. Key requirements for this project include:

Metadata management, data quality management, and system management services;

Sandbox application management;

Data mining, querying, reporting, and multidimensional analysis for self-service analytics;

Data computation and analytical visualization for non-performing loan governance;

Efficient data transfer with the primary data warehouse and Hadoop platform.

The Deep Analytics Cloud is a core part of the centralized business analytics system at China Mobile Headquarters, responsible for historical data storage and in-depth analysis. It consolidates data from the business analytics central data warehouse, the B (Business), O (Operation), and M (Management) domains, as well as internet analysis results and sampled data from the Hadoop cloud. This platform supports various open analytical environments, enabling rational allocation of data space and computing resources. Data distribution follows a more logical multi-tier structure, establishing lifecycle management mechanisms for data and applications with scientific deployment planning.