User Voice: China Mobile Big Data Platform—Practical Insights

Recently, the 25th China International Software Expo · China Database Industry Summit, hosted by GBASE, was held in Tianjin. At the summit, Wang Xiaoyu, Database Kernel Architect at China Mobile Information Technology Co., Ltd., delivered a keynote speech titled “China Mobile Big Data Platform—Practical Insights.”

The speech outlined the evolution of China Mobile's big data technology architecture and analytical database architecture, shared how the Wutong Big Data Platform is evolving from a lake-and-warehouse coexistence with coupled storage and compute to a cloud-native data lake foundation with a one-lake-multi-cloud architecture, and presented best practices for addressing challenges in cost, scalability, data silos, data migration, and cross-cloud collaborative management.

With the growth of the digital economy, China Mobile's data scale has exploded, data forms and types have become increasingly diverse, and various data applications have grown more extensive, penetrating the entire chain of internal services, production, operation, and management. At the same time, traditional data warehouse and big data platform technologies have gradually shown their limitations, tending to create data silos, high costs and low efficiency in data migration and sharing, and high barriers for data development, governance, and algorithms.

Given this situation, next-generation big data and data warehouse architectures such as cloud-native, storage-compute separation, and data lakehouse have been proposed. China Mobile has addressed cost and scalability issues by adopting a storage-compute separation architecture, and solved data silo and migration problems through a unified data foundation with lakehouse integration. It is gradually evolving from a lake-and-warehouse coexistence with coupled storage and compute to a one-lake-multi-cloud architecture with the data lake as the foundation and a cloud-native data warehouse as the engine, solving key issues in cost, scalability, data silos, data migration, and cross-cloud collaborative management.

The data warehouse of the big data platform is planned with five layers: interface layer, detail layer, asset layer, service layer, and application layer. Based on data processing flows and requirement characteristics, four data warehouses have been constructed, with a multi-warehouse, multi-cluster architecture designed. Data from the basic data warehouse and asset data warehouse is uniformly stored and managed by the data lake.

As a long-term partner of China Mobile, the GBASE series of databases has played a vital role in the construction of China Mobile's big data platform.

The distributed logical data warehouse GBase 8a MPP Cluster has enabled full-data-warehouse cloudification across the big data domain, splitting Hadoop clusters by business logic and providing transparent access capabilities for efficient data flow. Through application practices, it has validated the storage-compute separation architecture's ability to support China Mobile's business operations. By leveraging lakehouse integration for unified data collection and use, it has significantly improved model processing efficiency and greatly reduced both storage and computing costs.

The cloud-native data warehouse GCDW serves as the foundation for the big data platform to achieve data lakehouse unification, using S3 and HDFS to build unified data storage and providing elastic computing clusters of any scale.

Architecturally, it achieves storage-compute separation, stateless elastic scaling of service and compute nodes, simultaneous access to multiple heterogeneous storage systems, and cross-storage federated queries. In terms of performance, kernel optimizations have enabled vectorized execution engines, C++ native HDFS access, data caching, and operator pushdown. At the ecosystem level, it supports multiple cloud infrastructures, as well as deployment modes including cloudification, virtual machines, and physical machines, and has successfully adapted to a wide range of Chinese servers and operating systems.

Looking ahead, China Mobile will advance its architecture in two dimensions: unified storage and a unified data warehouse for the big data platform. It will work with GBASE and other partners to build a unified data lakehouse ecosystem for big data innovation, maximizing the value of data assets and applications.