Chinese-Built Database Empowers China Mobile's Big Data Platform

China Mobile Big Data Platform Background

 To effectively support China Mobile's “Big Connectivity” strategy, implement IT reform deployment, and establish a large IT system with “advanced architecture, efficient operations, and both internal and external balance,” China Mobile founded China Mobile Information Technology Co., Ltd. Internally, it operates and supports centralized IT systems; externally, it provides big data and other IT solutions and services to enterprises in other industries and overseas telecom carriers.

In the big data era, the construction of China Mobile's big data system faced common pain points: difficult data integration, single data services, challenges in building a big data ecosystem, low data processing efficiency, low-level data utilization, and incomplete management processes.

After passing over 30 indicator tests, GBASE's independently developed analytical database GBase 8a MPP Cluster was selected through China Mobile's evaluation criteria, demonstrating excellent performance in standard SQL support, scalability tests, and high availability. A total of thousands of GBase cluster nodes were deployed, with storage capacity exceeding 10 PB. This meets the needs for long-cycle data storage and cross-domain data integration, enabling the analysis and mining of massive data. It provides data and analytics services to group tenants, province tenants, and partner tenants while ensuring data service security.

China Mobile Big Data Platform Deployment

Centralized Deep Analytics Cloud Project Introduction
 

The centralized business analysis system initiated its first phase in 2014, divided into three sub-projects: the integrated application platform, the deep analytics cloud platform, and the Hadoop cloud platform. Key applications were progressively launched in early 2016, and the overall system was fully live by early 2018. The deep analytics cloud platform was built to ingest and provide long-period storage and management of data from the B, M, and O domains. At the application level, it supports both common and personalized needs of headquarters and provincial subsidiaries. Relying on the self-service analytics platform, it creates a comprehensive self-service analytics environment for various departments, pioneering new management and service models for business support.
 

The data warehouse of the Centralized Business Analysis System Deep Analytics Cloud Platform at China Mobile headquarters was constructed. By deploying application components such as query, reporting, multidimensional analysis, and sandbox management, it obtains data from various domains to form long-period historical storage of enterprise data. Supplemented by sampled data and sandbox environments, and with the help of self-service analytics tools, environments, and necessary models and application components, it provides self-service or customized deep services for data analysis across different departments.
 

As a data storage platform, the deep analytics cloud consists of four types of data: converged data from the B, O, and M domains; unstructured data, i.e., internet analysis and aggregated results from the Hadoop cloud; sandbox data and sampled verification data provided by self-service analytics; and data from data marts.

China Mobile Big Data GBase 8a MPP Cluster Scale

China Mobile deployed a total of thousands of GBase cluster nodes across its production center, disaster recovery center, and partner center, with storage capacity exceeding 10 PB. This meets the needs for long-cycle data storage and cross-domain data integration, enabling massive data analysis and mining, while providing secure data and analytics services to group tenants, province tenants, and partner tenants.

GBase 8a MPP Cluster Usage

Cluster Load: The production GBase clusters show high CPU and memory utilization. To further expand the scope of tenant support and data ingestion, and to enhance long-cycle data storage and deep integration, the GBase cluster is being scaled out as planned. Four major business zones were established to support data analysis needs from departments such as marketing, information security, internal audit, and government-enterprise business.

Centralized Onboarding of Provincial Tenants

In 2019, China Mobile Information Technology Company will continue to build a centralized, open, and shared big data cloud platform. It provides full-process, end-to-end, one-stop professional services for onboarded provincial companies, specialized subsidiaries, and partners, fostering a flourishing ecosystem of big data applications and jointly constructing a collaborative big data service ecosystem to accelerate the company's centralized strategy.

In 2018, the first batch of pilot onboardings covered four provinces including Tibet and Tianjin. In 2019, the second batch of pilots included seven provinces such as Xinjiang, Inner Mongolia, Gansu, and Hebei, with further expansion to other provinces planned subsequently.

Unified Management of GBase 8a MPP Cluster Tenants via PaaS Platform

To create a new paradigm where China Mobile's enterprise-level provincial big data platform is open to both internal and external customers, and to achieve the goal of supplying PaaS platform services externally, the platform leverages China Mobile's data sharing advantages and adopts an internet-oriented mindset to provide open resource services, tool services, and data services. As a service provider, GBase supplies tenant resource ordering, modification, unsubscription, and user management services in the form of multiple logical groups.

Value Delivered

A New Tenant Model: The GBase 8a MPP cluster will deliver services as a resource pool. Each provincial subsidiary is managed as a unified tenant, forming a centralized data aggregation and management paradigm. The MPP cluster can be scaled on demand based on the number of tenants, with cluster resources expanding linearly as provinces onboard.

Hybrid Architecture for Massive Data: Through distributed computing and storage, and a hybrid architecture incorporating Hadoop, MPP, and the master data warehouse, massive data is effectively supported.

Dynamic Scaling: The system offers strong scalability, supporting dynamic cluster expansion with performance that linearly improves as nodes are added.

Low Investment, High Efficiency: GBase 8a MPP Cluster runs on cost-effective x86 PC servers, delivering high performance at a low cost.

Domestically Developed: The product is fully independently developed, ensuring Chinese-owned autonomy and control, aligning with national development trends.