In recent years, China Mobile has been strengthening the independent and controllable construction of IT support system for business operation. AntDB has been cultivated in the core system of carriers for 15 years, and has been put online in many fields of the communication industry and guaranteed continuous and smooth operation, which is fully equipped with the capability of foreign databases. Because of its leading technology and stability, Hunan Mobile(the branch of China Mobile in Hunan Province) has been actively promoting the localization of its core business database through AntDB.

Ⅰ

Construction Requirements

Distributed architecture for massive data

South China Center is the key site of Hunan Mobile, and its detailed ticket database of billing system carries many types of business bills of mobile communication users, including traffic, voice, SMS, MMS and Monternet-type tickets, which are processed by the database after collection, billing, accounting and other business modules.

The number of communication users served by South China Center reaches tens of millions, and its detailed tickets database has very high requirements for data storage, analysis and calculation, and the data volume can reach billions of records added daily, and the designed capacity of the platform reaches hundreds of terabytes. The traditional centralized architecture cannot meet the business requirements in terms of concurrent performance, high availability and scalability, so the horizontal scalability of AntDB is needed to enhance the capacity of its system to carry the ticket volume and analyze and process the data.

The billing project team of South China Center adopted a gradual approach with reference to the scope of impact on business, starting from the Center’s billing system to gradually complete the replacement of multiple core databases.

The project team communicated in depth with the customer and analyzed the main needs of the customer as follows:

1. The domestic database needs to have a high degree of compatibility with the previous foreign commercial database, to minimize the  upper-level business changes, accelerate the replacement cycle and reduce the cost.

2. High-load business bearing. The daily processing volume of the database reaches several billion.

3. Massive data processing. Hundreds of terabytes of data are generated per month according to the existing business model, requiring database processing.

4. The architecture of distributed database cluster is complex, involving many node types and nodes, it will greatly enhance the complexity of operation and maintenance, reduce system reliability if merely relying on human operation and maintenance. So introducing a  intelligent operation and maintenance management platform of database is needed.

This project has the following requirements for database considering customer’s concerns:

 High compatibility with the previous database in terms of stored procedures and syntax

 Stable product functions, improved performance, and flexible expansion

 Highly available and disaster-tolerant architecture

 Full-lifecycle management and maintenance tools and related services

Ⅱ

AntDB Solution

Six advantages facilitate stable operation

  AntDB meets all the requirements of the customer, and it leads the industry especially in syntax compatibility with Oracle. In addition, what AntDB provides is more than a product, but a complete set of service for data management system, including a set of core product, a set of toolkit, intelligent operation and maintenance system and so on, which are convenient for its customers to manage the digital assets of the corporation in a quick and efficient way.

Figure 1 : product system of AntDB 

AntDB has six advantages as follows:

Figure 2: advantages of AntDB 

Highly compatible with foreign commercial databases

AntDB has the original multimodal SQL parsing engine: 1. customer can customize and expand the design of multi-syntax SQL parsing engine, which is compatible with the ecology, and one set of database can meet the requirements of localization and replacement of various databases including foreign commercial databases of this project, reducing the learning cost of customers. 2. The switch of SQL parsing engine takes effect dynamically without restarting the database. 3. It supports three access modes of session mode, service mode and statement mode, offering the ultimate experience of flexibility and ease of use.

The high compatibility with foreign database syntax: syntax compatibility, function compatibility, system table/view compatibility, stored procedure and trigger compatibility, data type compatibility, OCI and JDBC call interface compatibility, management package compatibility, rowid/rownum and dual dummy table, etc. Based on the feature of high compatibility, the heterogeneous database disaster recovery of AntDB and foreign databases can be realized, and two sets of databases running in parallel at the early stage of launch to reduce customer's doubts.

Cloud-native distributed architecture and elastic scalability

The horizontal scalability of database is required for detailed ticket library of settlement center to realize linear expansion. With cloud-native distributed design, AntDB is completely transparent to application and perfect in performance and scalability. In addition, AntDB has powerful scale up capacity, which holds no limits to CPU, memory and the number of connection. When the data is sliced and stored in each data node, the number of nodes can be expanded horizontally to enhance the database storage capacity by virtue of the distributed architecture when the data storage demand of detailed ticket library exceeds the limit of a single server.

Real-time analysis of data

The billing and detailed ticket database processes billions of tickets per day, and performs data analysis to generate reports. AntDB has real-time strong consistency of data and distributed transaction control, ensuring 0 data loss and transaction consistency, and guaranteeing the accuracy of bill data processing. AntDB has dynamic and efficient SQL execution engine, providing heterogeneous index and built-in reduce data processing engine, and data nodes can communicate with each other to improve data processing efficiency, especially for complex query scenarios (aggregation, correlation, etc.), and improve the processing efficiency of billing statement data and the generation speed of related report.

Figure 3: the architecture of AntDB 

High availability and load balancing

The billing system requires high reliability. AntDB provides a complete cluster self-healing solution to ensure second-level automatic failover to achieve business continuity, and can guarantee data integrity and strong consistency to achieve the true sense of Auto-Failover, ensuring that the database system can obtain 99.999% SLA service guarantee.

AntDB adopted typical distributed architecture for the South China  Center, which has dual-center disaster recovery mechanism with total 48 hosts in master and standby center. The data nodes DN1-DN12 are equipped with one host and two standby machines each, where the master center is configured with one host and one standby machine while the disaster recovery center with an asynchronous backup node, totaling 36 nodes. The GTM master-standby center has one host and one standby machine for each one, totaling 4 nodes. The master-standby center of compute node has 4 machines for each, amounting to 8 nodes, and the data between master and standby node is synchronized through stream replication. The data nodes from DN1 to DN2 are master nodes, and the rest of them are standby nodes. The nodes in standby data center synchronize the data by cascade replication. Switch operation is completed through the command of ADBMGR. The architecture is as follows as shown in Figure 2 below.

Dual-center disaster recovery ensures reliability

According to user’s requirements, the billing system is the core system, and to ensure the business continuity and data security, the database needs to build a main/backup disaster recovery system, so that another center can quickly take over when a global failure occurs in a single center, such as power failure or network disconnection, without affecting the data storage and processing of biling tickets. AntDB supports various disaster recovery methods, such as dual-center, two locations and three centers, etc. The dual-center disaster recovery mode is chosen for this project.

AntDB distributed deployment architecture naturally supports dual-center disaster recovery, and through AntDB cluster management tools, the cluster can be quickly built in the backup center, and can be quickly switched to the backup center to carry business when the global failure occurs in the main center.

Figure 4: the "dual-center" deployment architecture of AntDB’s distributed disaster recovery

Intelligent operation and maintenance control

The number of nodes of billing and detailed ticket database in the is extremely large, and the operation and maintenance platform needs unified management of database clusters. AntDB provides intelligent operation and maintenance control platform, which can realize management of database cluster. Automatic monitoring of database, including regular database server monitoring, database key indicator monitoring, and alarm notification, etc., can greatly improve the maintainability of database.

Figure 5: the intelligent operation and maintenance control platform for database AntDB Control Center(ACC)