GBase 8a MPP supports the data layer of a provincial China Mobile cloud-based business analysis system

The big data foundation platform for the data layer of a provincial China Mobile cloud business analysis system uses GBase 8a MPP Cluster as the data warehouse and data mart platform to integrate data from various business systems, delivering the following platform value:

1. Low Cost
● The GBase 8a database runs on low-cost x86 PC servers, effectively reducing hardware investment.
● The overall cost of the new system is reduced to approximately 1/10 of the original.

2. Elastic Scalability
● The GBase cluster efficiently handles petabyte-scale data, meeting storage requirements for both base and application data.
● Shifting from a traditional vertical scale-up model to a horizontal scale-out model based on data volume enables dynamic scaling without service interruption, ensuring continuous service availability.

3. High Availability
● Through proper configuration, efficient load balancing is achieved, fully leveraging each node’s computing capacity and improving overall system synergy.
● Backup strategies based on security groups ensure that node failures do not disrupt service continuity.

4. High Performance
● Enhanced ability to deliver external data services enables data monetization and revenue generation.
● Efficient data computing and query performance meets the demands of various specialized analyses and innovative applications.

1. Overall Architecture of the Cloud-Based Business Analysis System at a Provincial China Mobile Branch

Figure | Overall Architecture of the Cloud-Based Business Analysis System at a Provincial China Mobile Branch

The data layer of this cloud-based business analysis system consists primarily of a cloud-based historical data store, a big data foundation platform, and a big data analytics platform. The big data foundation platform is built on a distributed MPP database (GBase 8a) using x86 technology, serving as the data repository for the main business analysis warehouse and the big data analytics platform. It carries massive historical data mining, trend analysis, and forecasting workloads. The platform is composed of a main data warehouse and data marts. The main data warehouse processes core data models and supports a limited number of high-timeliness applications, while the data marts draw base data from the warehouse and deliver end-to-end applications.

2. GBase 8a MPP Main Data Warehouse

Figure | Main Warehouse Database Network Architecture

The data warehouse was deployed as a new MPP cluster managed by the GBase UP platform, which converges MPP and Hadoop heterogeneous engines. The MPP cluster carries all original main data warehouse workloads (including first-level and provincial business analysis). The Hadoop cluster stores cold data from the historical data store and provides query access through a transparent gateway. All data exchange between the MPP cluster and external systems is handled by the cloud-based ETL platform and the BDI data exchange platform. The DACP platform provides unified management of data, security, tasks, and scheduling on the MPP database.

Figure | MPP Cluster Deployment Diagram

Figure | Data Flow Diagram

Data from the BDI data exchange platform is cleaned, transformed, and filtered before loading into the GBase cluster, with data, security, tasks, and scheduling managed via DACP.

The deployed MPP cluster uses SSD disks to boost read/write performance of the main data warehouse. It runs GBase UP, leveraging a new MPP + shared-nothing distributed architecture that delivers significant improvements in node scalability, data loading performance, high concurrency, high availability, and online expansion. In particular, it adds support for large-scale structured and unstructured data processing technologies from the Oracle and Hadoop ecosystems, further addressing the evolving demands of business and operations.