ITPUB Interview | Diving Deep: What Are the Core Requirements for OLTP Databases?

In the past two years, the drive to adopt domestic databases has entered a deeper phase, accelerating the modernization of core systems in critical sectors such as finance, telecommunications, and rail transit. According to industry reports, centralized databases still dominate the technical architectures used in these key industries. The Financial Industry Database Supply Chain Security Development Report (2022) found that centralized databases account for 89% of total deployments in the financial sector — 80% in banking and over 90% in securities and insurance. The Database Development Research Report (2023) published by CCSA TC601 Big Data Technology Standards Promotion Committee also indicates that centralized databases remain the predominant deployment model for the telecom industry.

As the adoption wave reaches deep waters, what challenges and opportunities lie ahead for domestic databases? Should organizations pursue a simple drop-in replacement or a comprehensive upgrade? In a thriving database market, what are the key priorities and hurdles for today’s centralized databases? Guo Zhuo, head of the GBase 8s product division at GBASE, recently discussed these topics in an interview with ITPUB, a professional database and architecture technology community in China.

Interviewee: Guo Zhuo, Head of the GBase 8s Product Division at General Data Technology. With more than a decade of hands‑on experience in the database industry, he has worked on the front lines of China’s database evolution, taking on roles in data development, technical support, and product management.

Q: What core systems really need from a database — how do you select the right one?

The replacement or upgrade of databases with domestic alternatives across industries has generally advanced from peripheral to core, and from partial to full-scale. Core systems place the strictest demands on stability, performance, high availability, and security. Guo Zhuo explains that core systems in different industries also have distinct requirements for transactional (TP) databases.

The financial and telecom industries focus primarily on data processing power, stability, data consistency, and high availability. For instance, core banking systems require strong consistency and high availability to ensure accurate service delivery and uninterrupted operation at all times. In telecom, massive volumes of data are generated daily, demanding that TP databases handle high concurrency pressure and provide efficient data processing for rapid transactions and data management.

Government and public sector organizations frequently handle sensitive information, such as national ID data, so they need databases with robust security features — for example, user authentication and authorization, and data encryption to safeguard privacy and security.

In rail transit, continuous high-frequency transactions — like fare gates or ticket counter transactions — place high demands on database stability, processing performance, consistency, and concurrency capabilities.

Overall, the cross-industry requirements for TP databases converge on reliability, data integrity and consistency, processing power, and data security. Guo Zhuo emphasizes: “None of these high standards can be met at 80%; they all require 100%.”

Requirements may be clear, but choosing the right database is no easy task. Guo Zhuo points out that organizations typically evaluate the following factors during database selection:

First, performance and stability — how well the database performs when processing large data volumes or high concurrency requests, and especially whether it remains stable over long-duration operations. High availability and fault tolerance are also scrutinized.

Second, compatibility. Many users migrating to domestic databases prioritize syntactic compatibility with Oracle, DB2, MySQL, and other established databases, as this directly impacts the overall cost of replacement.

Third, ecosystem support, such as community resources, documentation, and training.

Fourth, technical service and support — ongoing product iteration and bug-fix capabilities. Professional technical support from the original vendor is often more friendly for system operations teams and can reduce overall maintenance costs.

A truly effective database product is always shaped by real-world user scenarios. The industry broadly agrees that domestic databases are transitioning from “usable” to “well-made,” and achieving that next level is key to future success. What defines a well-made database may vary by perspective. From GBASE’s viewpoint, a well-made database is one that users can consistently trust.

Building a trusted TP database

GBASE, celebrating its 20th anniversary this year, is a veteran database vendor in China and a full-stack database provider. GBase databases span AP, TP, and HTAP solutions to meet diverse business scenario requirements.
“Many people are more familiar with our AP analytical database, GBase 8a. In fact, our TP transactional database, GBase 8s, has an even longer track record,” Guo Zhuo notes.

GBase 8s was first introduced in 2005, making it GBASE’s longest-standing database product. In 2014, aligned with the company’s strategic road map, we set our sights on covering OLTP scenarios across all industries and began a fundamental architectural redesign and upgrade of the 8s product. In 2016, the overhauled GBase 8s was formally launched, possessing the native high-availability clustering capabilities based on shared storage and the product strength needed to replace core OLTP systems across industries, with 100% independently controllable source code.

Since then, GBase 8s has entered a new phase of growth. Following the release of GBase 8s v8.8 in 2017, it gradually gained traction in key industries. As its ecosystem matured, it steadily strengthened compatibility with Oracle, MySQL, and other mainstream database syntaxes. Today, GBase 8s has become an OLTP database that supports hundreds of terabytes of data, with high security, high reliability, and rigid transactional consistency.

To meet the varying requirements of different business systems, GBase 8s offers flexible high-availability solutions, including: shared-storage clustering (SSC) or primary-standby high-availability disaster recovery (HAC) within the same data center, with RPO=0 and RTO < 10 seconds; intra-city primary-standby disaster recovery HAC with RPO=0 and RTO < 30 seconds; and remote disaster recovery (RHAC) with RPO measured in seconds and RTO < 2 minutes. Failover occurs within seconds, and the overall high-availability architecture achieves 99.999% uptime. All high-availability capabilities are built directly into the database service itself without relying on any third-party components, enhancing both stability and reducing deployment and maintenance costs.

Today, GBase 8s can replace international mainstream databases in over 80% of scenarios and replace Oracle and Oracle RAC clusters in more than 95% of scenarios. It is widely deployed in core systems of government, finance, telecom, rail transit, energy, and many other critical sectors, standing as one of GBASE’s three flagship products. Its cumulative deployments exceed 50,000 nodes.

For example, in a large central-government-owned enterprise project, GBase 8s provided a three-site-two-center high-availability architecture. The project went live in 2017 and has since accumulated nearly 100TB of business data, supporting hundreds of application microservices. Daily TPS routinely exceeds 2,000 — meaning that what would normally be peak workload for an ordinary business system has become the new normal, placing immense pressure on the database. The system’s longest continuous operation stretch has surpassed 600 days, a clear testament to GBase 8s’s stability and resilience.

In the rail transit industry, GBase 8s has achieved full coverage, powering independent, controllable, and security-trusted core business systems such as line network command centers, clearing systems, ticketing systems, multi-line center systems, and integrated monitoring systems. Since breaking into the rail sector, GBASE has rolled out its solutions across more than 20 cities — including Shenzhen, Chongqing, Xi’an, and Chengdu — spanning numerous metro lines. It supports the efficient and stable operation of core systems for roads, railways, aviation, and maritime transport. Shenzhen Metro, in particular, faces passenger volumes, data volumes, and processing demands that rank among the most intense in China. GBase 8s has been deployed on Shenzhen Metro and has run stably for extended periods, earning genuine trust from the customer.

Q: Drop-in replacement or upgrade? How to achieve a smooth migration?

The wave of database localization has opened up opportunities, and two strategies dominate: straightforward replacement (“lift-and-shift”) and comprehensive upgrade. Which to choose, and how to migrate smoothly?

Guo Zhuo believes that both approaches have their own advantages and challenges, and the choice should be based on a comprehensive assessment of business needs, data complexity, and migration difficulty. In general, smaller-scale business systems are better suited for a direct replacement, while large-scale core systems may benefit more from an upgrade path. To ensure a smooth transition during database localization, she offers the following recommendations:

First, thorough testing. Before going live, conduct full testing in the production environment — functional, performance, and stress testing. It is also advisable to perform destructive testing to verify the end-to-end performance and stability of the new business system, from hardware to software.

Second, phased migration. Depending on business realities, a gradual migration can be chosen. Long-running, large-scale business systems often have extensive historical data. A cold/hot data phased migration to the new database combined with a dual-track parallel operation model can achieve a more reliable cutover through flexible migration between old and new environments.

Third, backup and rollback strategies. Ensure comprehensive backup strategies and rollback plans are in place to address any unexpected issues. Typically, before a system goes live, the GBase 8s team customizes automated backup scripts and provides general rollback solutions according to each system’s specific requirements. For example, during the core system replacement at Luzhou Bank, the GBase 8s expert team tailored a rollback plan that ensured minimal business disruption by enabling a rapid rollback if problems occurred during the upgrade cutover.

Fourth, training and support. Provide sufficient training and support so that after project delivery, operations and development teams can proficiently use the new database system. GBASE offers product training before every system go-live, and has established a three-tier certification system to help customers nurture professional technical talent.

Fifth, monitoring and optimization. Whether you choose a drop-in replacement or an upgrade, post-go-live monitoring and continuous business optimization for the new database are critically important.

As the saying goes, just as water retains no constant shape, in warfare there are no constant conditions. Guo Zhuo emphasizes that the final decision must be made case by case. Each business system has its own requirements, and the priorities of the suggestions above may vary from one organization to another.

In closing

To better support core systems and seize the opportunities of database localization, GBase 8s has mapped out a clear future direction. It will continue to strengthen syntactic compatibility with mainstream databases such as Oracle, enhance overall ecosystem development, provide multi-write multi-read capabilities, support multi-version concurrency control (MVCC), enable flexible on-premises and cloud deployment, and adopt RDMA high-speed networking.

A truly great database is forged with craftsmanship. “Despite the uncertainties we face today, one thing is absolutely certain for GBASE: we will keep focusing on building enterprise-grade database products that our users can trust.”