Lao Yu Interview Series (III): Decoding China’s Database Shared-Storage Clusters and Sync Tools

Introduction

As database technology evolves rapidly, shared-storage clusters are becoming the ultimate choice for finance, government, and other core scenarios. Recently, the well-known media outlet Lao Yu’s Notes sat down with Cui Zhiwei, General Manager of the GBase 8s Product Department at General Data Technology, to discuss the GBase 8s database. Drawing on real-world cases, the two experts explored how shared-storage clusters achieve breakthroughs in both business continuity and data security through dynamic scaling and optimized failover. Addressing industry skepticism about insufficient database ecosystem tools from Chinese database vendors, the dialogue also revealed how GBase 8s uses its modular operations tool GM to shatter the “black box” image and make its technical strengths intuitively empower business.

Lao Yu: GBase 8s has been running stably in the Shenzhen Metro fare clearing system for more than two years. In the materials I’ve seen, it mentions something quite interesting—that shared-storage cluster technology has become the preferred choice for core banking applications. For users who have adopted GBase 8s’ shared-storage cluster, what specific improvements have they seen in transaction processing efficiency and data security?

GBase 8s Cui Zhiwei: In fact, for any centralized database, clusters are indispensable. In the database cluster world, what we typically see from open-source options are primary-standby clusters—one primary and one standby, one primary and two standbys, or even one primary with three standbys. My view is that the ultimate destination for a centralized database is the shared-storage cluster. First, its scalability is truly flexible: computing nodes can be dynamically added online, and storage can be scaled independently. Second, within certain business scale ranges, its total cost of ownership (TCO) is optimal. Just look at Oracle RAC, or even some Chinese databases today—everyone is moving in this direction. What does that tell us? It shows that this is a real trend, one that’s been validated in practice.

Let me give you a concrete example. GBase 8s has been running stably in the Shenzhen Metro fare clearing system for over two years; that case speaks volumes. There’s also the Luzhou Bank case. After they adopted the shared-storage cluster, their most immediate feedback was that transaction processing efficiency went up, and data security improved significantly. To put it simply, the biggest advantage of this technology is “dual dynamic scaling”—you can expand computing nodes online and scale storage independently, just like building blocks.

Lao Yu: When it comes to fault handling, what key optimizations has GBase 8s focused on this year?

GBase 8s Cui Zhiwei: This year we made a major improvement in failover. In the past, when a failure occurred, it was like downloading a large file and suddenly losing the network connection—you’d have to start the whole download over. Now, we’ve implemented what you might call “resumable data transfer” failover. For example, suppose you’re deleting 50 million rows of data, and at 49 million the node suddenly fails. Traditionally, you’d have to roll back and start over. Now? After switching to the standby node, you only need to continue with the remaining one million deletions. As a result, business interruptions last just a few seconds, and the boost to business continuity is truly substantial.

Lao Yu: One common complaint about Chinese databases is the lack of ecosystem tools—people often fall back on open-source tools. I’ve heard that your GM tool is the O&M weapon behind GBase 8s. What was the original design logic? What problem were you trying to solve?

GBase 8s Cui Zhiwei: When we first sold GBase 8s database, we basically just provided a server plus two driver packages. We quickly learned that wasn’t enough—customers couldn’t really perceive our technical advantages. Now, with the revamped GM tool, we’ve adopted a plug-in design, modularizing resource center, monitoring & alerting, and O&M management. This actually makes things quite flexible: customers just mix and match plug-ins like building blocks—they install only what they need. We’ve seen this flexibility work well with metro systems, the Bank of Guangzhou, and other end-users. Whenever databases are heavily used, they all reach for GM, which visualizes applications very intuitively.

Lao Yu: There was a comment on one of my earlier articles that I found particularly interesting. The commenter wrote something like, “Isn’t there only one company in China that has RAC? Does GBase also have it?” Many people have no idea that GBase 8s offers a shared-storage cluster. So I think you should promote it more and talk about it more often.

GBase 8s Cui Zhiwei: I can really relate! Earlier this year I visited a centrally-administered state-owned enterprise for a technical exchange. When we mentioned that we have shared-storage clusters, primary-standby clusters, and more, the other side was stunned: “We’d heard that only one company in China has such clusters.” At the time I said, it’s normal for everyone to promote their own products. But let’s be realistic—since what year did our shared-storage cluster go live, and what production system is it running? We have genuine cases. However, we definitely do need to step up our promotion.

Lao Yu: Having reported on databases for many years, I know that databases are quite low-level and very technical. How do you make such foundational database technology understandable to non-technical audiences?

GBase 8s Cui Zhiwei: That’s indeed a big challenge. We mainly approach it in two ways: first, through the GM tool, we make technical capabilities visible, so operators can see the database’s running status intuitively. Second, we use everyday analogies, like “resumable data transfer” to explain failover, so that business stakeholders can grasp it more easily. After all, technology must ultimately serve business—it should never become a black box.

Lao Yu: In which directions will GBase 8s continue to strive in the future?

GBase 8s Cui Zhiwei: We’ll focus on two main areas: continuously improving the high availability capabilities of our shared-storage cluster, and deepening the intelligent O&M features of the GM tool. Most importantly, we will keep listening to customers’ actual needs, using technology to solve real business pain points. Because at the end of the day, the quality of a database is measured by whether it creates tangible value for the customer’s business.