Project to Replace the Database in the Network Management System of Jiangxi Mobile with an Independently Developed Database

Project Background

The mobile network management (NMS) system is designed to allocate and manage network resources and information, monitor network operations in real time, and dynamically adjust resources based on service demands and traffic usage.

As Jiangxi Mobile's business grew, the NMS system's processing load continued to increase. The system relied on an outdated foreign database version, and recurring product issues frequently caused operational interruptions. Moreover, China's IT strategy emphasizes technological self-reliance, gradually replacing foreign software and hardware with domestic alternatives. This project aimed to migrate the NMS database to a network service resource pool.

Challenges and Requirements

Since its launch in 2005, Jiangxi Mobile's NMS system lacked a high-availability architecture. Although hardware had been expanded over time, CPU utilization frequently reached 60-70%, memory usage 70-80%, and storage capacity was critically limited. The system could not add database logical logs, often leading to long-running transactions, and there was no backup space for full backups. Key issues included:

• The system ran on a single server, posing a significant single point of failure risk.

• Hardware resources and performance could not meet current or future business demands.

• The outdated database version frequently caused operational disruptions.

• The large volume of business data hampered system performance and backup operations.

• The legacy minicomputers incurred high maintenance costs, with slow support service response.

Project Requirements

The project requirements were as follows:

• Eliminate the single point of failure by deploying a dual-server high-availability cluster to ensure overall system resilience.

• Upgrade hardware resources and performance by replacing the underperforming minicomputers with more powerful PC servers.

• Adopt a mature, domestically developed and controllable platform to reduce operational and maintenance costs.

• During the system migration, clean up historical data to streamline the existing production environment.

• The new system must seamlessly take over existing services and accommodate future business planning.

• The migration must not disrupt ongoing operations, delivering a seamless cutover.

Project Solution Highlights

Replace existing outdated foreign minicomputers with high-performance PC servers, and replace old foreign storage with higher-performance, larger-capacity storage;

Replace the old foreign commercial database with the mature and stable GBase 8s database;

The entire business system adopts a dual-node architecture, leveraging GBase 8s Shared Storage Cluster (SSC) technology to eliminate single points of failure and enhance overall system high availability;

• When future business growth occurs, the system can be scaled out by adding database cluster nodes to meet evolving demands;

• During the replacement, historical data in the original system will be cleaned up to streamline the existing production system;

• During the replacement, resource planning for new business functions will be allocated to ensure the long-term stable operation of the new system;

• Given the repeatable nature of network management system applications, the old system will continue to run while the new system catches up on data. After synchronization, both systems will run in parallel for 3 months to guard against any unforeseen issues with the new system and to allow time for optimization.

Solution Architecture Diagram 

In the new system, read-write network management operations are routed through the Connection Manager to the primary node, while read-only operations are handled by SSC secondary nodes via the Connection Manager. Beyond read/write splitting and load balancing, the Connection Manager also quickly promotes a healthy secondary node to a writable new primary if the primary fails. Failover time depends on the load at the moment of failure, typically completing within 20 seconds under moderate loads. To ensure overall high availability, multiple Connection Managers are deployed and separated from database servers to avoid interference. In addition to a heartbeat network connection for health checks, the system also implements disk heartbeat—allowing the primary and secondary nodes to communicate through shared storage—to prevent split-brain scenarios if the heartbeat network fails.

This project involved replacing foreign products with homegrown alternatives. After testing and operation, the replacement was completed smoothly with remarkable results. It not only delivers clear cost advantages but also boosts performance and the ability to handle large-scale data. This fully demonstrates that GBase8s SSC technology can completely replace the RAC functionality of foreign commercial databases.

Summary of Application Results

• High Stability

The new system has been running stably for over a year without any failures.

• Fast Failover

During the parallel running of the old and new systems, we simulated production system hardware and software failures including network and power outages. The production system performed a dual-system failover within 20 seconds, continuing to provide services and meeting business requirements.

• Load Balancing

The secondary node of the dual-system setup can handle more query functions through the database connection manager, achieving load balancing for the database cluster.

• Scalability

As the business grows, the system can scale out by adding more cluster nodes to meet future business development.

Key Value Delivered

• With a Chinese-developed database, the technical support service is far superior to that of foreign products.

• After replacing the database and other software and hardware, the overall system performance has improved significantly, and the system is much more stable.

• The new system adopts a dual-system high-availability architecture with shared storage, preventing the risk of single-node failures.

• Businesses that previously ran on a single machine can now run on both nodes, facilitating business expansion and allowing hardware resources to be fully utilized.

• As the business continues to grow, new business loads can be supported by adding cluster nodes.