Titanium Media Exclusive Conversation with Yang Bing, CEO of OceanBase: After independent operation, change and change

As the self-developed database behind 11.11, OceanBase, which was born in 2010, is ushering in a key development node.

In 2020, OceanBase was established and commercialized, with only 18 customers that year. At the end of 2023, OceanBase announced its latest commercialization progress: serving more than 1,000 industry customers, with a 150% increase in customer growth over the years, of which 30% of customers apply it to core systems.

On March 19, Ant Group officially announced that its subsidiaries Ant International, OceanBase, and Ant Digital have established a board of directors and began to independently face the market.

Recently, Yang Bing, CEO of OceanBase, had an exclusive conversation with Titanium Media App.

During the conversation, Yang Bing explained in detail the organizational formation adjustments made by OceanBase in order to better participate in market competition: the establishment of the Finance & Government and Enterprise Business Department to further support proprietary cloud customers by concentrating its advantages, the establishment of the Public Cloud Business Department to increase investment in public cloud and strive to create 1-2 large-scale replication effects in some specific directions, the establishment of an international business department to explore overseas expansion, and the merger of expatriate domestic employees and overseas local recruits to form a new overseas team.

In terms of development route, Yang Bing emphasizes the value of "root self-research". Self-development means that we must go through a long development cycle, from 0 to 1 self-developed 3 million lines of code, to the long-term cultivation of open source ecology and talent ecology, to the arduous challenge of productization and commercialization of technology, every step is a challenge to cross no man's land, and it is also a process of cultivating oneself and building barriers. Yang Bing told Titanium Media App, "Even so, we are very sure of the technical direction of root self-development, and we have no better choice for the new challenges of modern application architecture." ”

Talking about structural adjustment: finance, government and enterprises are the basic market, and new growth points are invested

TMT: What does it mean for OceanBase to operate independently?

Yang Bing: Independent operation is the second acceleration of OceanBase, before we went from 0 to 1, and in the future, we will go from 1 to N, accept the test of stronger market, and achieve a higher growth rate.

For example, OceanBase is a car that has completed the inner lap and reached the outer lap, and only by running in the top few places on the high-speed track in the outer lap can it become F1.

Titanium Media App: What adjustments have been made to the product or organizational structure of this independent market?

Yang Bing: We have made three adjustments. First of all, in the financial, government and enterprise fields, where we already have experience and advantages, we will merge the original carriers, financial and other industries, and the corresponding delivery teams into a large business unit, concentrate superior resources to further serve proprietary cloud customers, maximize the use of resources, and make the decision-making of the business department more efficient.

Second, we strategically invest in the public cloud and explore the second growth curve of OceanBase. Merging teams that focus on the public cloud industry can efficiently accelerate in the public cloud field, and also help some teams that can form a large-scale replication effect quickly get resources.

Thirdly, we send business students with pioneering experience in China to merge with overseas local teams to form a new overseas team to accelerate the replication of domestic experience and accelerate the process of overseas expansion.

Generally speaking, public cloud and overseas markets are our new growth points, while finance, government and enterprises are still our firm fundamentals. For distributed databases, China is the best cultivation market, with the largest massive data scenarios, and at the same time, it is also facing the opportunity of digital upgrading of a large number of core systems.

Titanium Media App: After OceanBase independently faces the market, it is bound to start pursuing a larger commercialization and market scale. In this context, where do you see future innovation?

Yang Bing: Innovation is mainly focused on integration. We will continue to practice the "all-in-one" product strategy, for systems with a certain degree of complexity and scale, for the operation and maintenance and management of the system, the unified technology stack of all in one will greatly simplify the operation and maintenance, and also make the system architecture easier to plan and unify, for application developers, through a unified programming interface or a set of SDK can handle different business loads, which will also greatly improve development efficiency and reduce team collaboration costs. Whether it is the convergence of TP and AP, or the convergence of programming models of NoSQL and other data types, the integrated convergence architecture and programming interface are our next strengths.

In addition, in the future, the storage media in the digital world will tend to be unified, and our goal is to build an integrated solution in this foreseeable future to achieve the unification of structured and unstructured data storage, so that OceanBase can become the most suitable data storage platform for modern application architecture.

Talking about the self-developed route: because of certainty, because there is no choice

Titanium Media App: Where does the confidence to choose the self-developed technology route come from?

Yang Bing: I think our confidence lies in two aspects.

First of all, our firm belief in the direction of technology is our confidence. Ten years ago, Ali Ant encountered a double 11 problem, which could not be solved by the original centralized database and open source technology, or the cost of solving it was huge. We firmly believe that only self-research can fundamentally solve the core problems of database application in the industry, and this belief also makes us dare to make decisions and assume the responsibility of solving problems.

Secondly, it is a deep understanding of the application of the scene. We believe that the biggest challenge of technology and products is not the difficulty of design or demonstration, but whether there are suitable scenarios for polishing and verification. Fortunately, we have our own products and business scenarios, which provide a natural testing ground for our technology development. This unique advantage gives us the confidence to overcome the challenges in real-world applications.

TMT: What kind of value does self-research bring to you?

Yang Bing: For two years, my DingTalk signature was "Building Bridges in Mountains and Waters", and R&D is such a process, OceanBase is made from 0 to 1, our talent pool and our engineering system are also tailored from 0 to 1, and the whole team has accumulated a very rich ability to deconstruct complex problems in the database field and strong innovation ability.

Hundreds of people working together to open millions of lines of code is a very complex system project, during which there will be a large number of bugs, in order to be able to fix such a system stably, it is necessary to have the ability to control and converge bugs in the project. We have a complete R&D process, testing system, security scanning system to ensure efficient and high-quality iterations, and we also have an internal practical application verification system to ensure production-level reliability.

The second is innovation, we can break the original architecture design and fundamentally solve the problem, including our LSM-Tree high-compression storage engine, the implementation of 2F1A in Paxos to achieve true three replicas, and the single-node distributed integration, etc., all of which are based on breakthroughs brought by self-development, which are actually difficult to achieve on the classic architecture of the past database, and can only be overturned and restarted. But only in this way can we redesign products based on first-principles principles, and can we break the routine and create greater value.

Talking about open source and closed source: Many people have misconceptions about open source

TMT: Regarding the difference between open source and closed-source databases, how do you view this difference, and how to find a balance between reference and innovation?

Bing Yang: I think a lot of people don't fully understand what true open source is. Based on the practice of open source but not giving back to the community, or the part that has been changed is not open source, I don't think it can be called open source. This practice may actually be closer to closed source, where the modified product is a black box. True open source should continue to contribute back to the community. Not only using open source technology, but also contributing your own improvements back to the community is a true understanding and support for the spirit of open source.

Another challenge for products based on open source modifications is the control or voice of future code branches, which requires huge investment and community influence, and even then there will be many disagreements caused by non-technical factors. Therefore, often products based on open source modifications will enter a more tangled situation, and if you go to the left, you may encounter the problem that you cannot feed back the changes to the upstream because you don't have enough right to speak, resulting in your own version becoming an orphan version. Although you can enjoy the dividends of the iterative development of the community, you can't build too deep competitiveness and answer the question of which abilities no one has. The former will move towards another kind of "self-development", and the latter will become a "shell" without differentiation.

OceanBase is a self-developed database that uses the open-source model of opencore. We've open-sourced the core of 3 million lines of code and forked the commercial kernel and keep it up to date. We enjoy the benefits of not having the baggage of historical code and design constraints, but we also need to bear the time cost of cultivating an ecosystem from 0-1. It's a choice, not a balance. Different paths may be parallel, intersecting, and similar in the short term, but they are different from the end. On the historical stage of the world, self-research is a road that requires huge investment but can be walked wider.

Talking about financial databases: the requirements focus on compatibility, distributed system maturity, and ecological adaptation

TMT: Please talk about the various factors that financial institutions consider in the process of database selection. How have their needs changed, and how do they differ between organizations of different sizes?

Yang Bing: In the financial industry, the needs and concerns of large institutions and small and medium-sized institutions are different.

The infrastructure of large institutions is better, and the focus is not only on distribution, but also on a complete unitized distributed solution and how to establish a highly available technical risk system under the distributed architecture. In addition, due to the large number of systems, large amounts of data, and diverse and complex infrastructures, large organizations are also concerned about the security and transformation costs of the entire solution, and also require higher compatibility of databases.

Small and medium-sized financial institutions, which prefer one-stop and integrated solutions, need the distributed capabilities of databases, but do not want to be perceived in use. The emphasis on compatibility and quick resolution of existing issues requires not only technical products, but also requirements for service capabilities and ecosystems, especially in terms of local service support.

TMT: What are the new developments of financial institutions in the process of iterative database upgrading, and what are the current new trends?

Yang Bing: The needs of financial institutions are still focused on compatibility, the maturity of distributed systems, and ecological adaptation. In addition, the support and adaptability of domestic chips are also the focus of attention. In general, financial institutions, whether large or small or medium-sized, are looking for technical maturity, strong service capabilities, and ecosystem support in the process of database selection and upgrade.

Talking about market competition: free is the most expensive

TMT: How do you view the price war in the database market?

Yang Bing: For a period of time, the database market competition was very fierce, and sometimes there was a price war. Some large manufacturers or cloud vendors may play a price war of "wool out of pigs", and the database software does not need money, but pays for other parts. Now that the market is more sensible, this phenomenon of free or low prices has decreased.

Price competition is a normal phenomenon, but a company with a database as its only core product and a company with a database as only one of the very important products must have an essential difference in the long-term price strategy. It's a choice, and of course I think price is also the basis for the product's strategic position in the business and its long-term sustainability. In addition, there will be a huge difference in pricing between a product-centric company and a company that focuses on overall solutions and projects, which will ultimately affect the sustainability of the product itself.

For OceanBase, we always focus on winning customers through the strength of technology and services. In fact, customers also understand that there is no such thing as a free thing, and free is the most expensive. As long as it is the core database of the enterprise and the database that loads the key business, expensive and free are not the most important.

Talking about the impact of AI: AI will hit the application layer first

TMT: Will AI have a particularly big impact on the database, and will it make you refactor the architecture?

Yang Bing: The impact of AI on the database is indirect, and AI will first have an impact on the application layer, and the change in application form will affect the database.

Over the past decade, the rapid development of the Internet has driven the pursuit of agility, servitization, and cloud computing. These requirements have led to the evolution of specific application forms and development models, which in turn have a fundamental impact on distributed system architectures. As a result, distributed databases have been widely developed, and I believe that distributed databases will become mainstream in the next few years.

In everyone's imagination, it may be similar to the search engine entrance of the past, such as Baidu search box or Google, or it may evolve into a dialog box, which will become the ultimate form of most applications. Now the development of AI is changing with each passing day, and one or several patterns have not yet been formed, so it is not clear what impact the demand for data storage software will have, but I am relatively sure of several trends:

1. Due to the dependence on computing power, AI as Service is an inevitable form, so cloud native will continue to prosper and deepen.

2. Unstructured data (video, pictures, and audio are mainly used in modern applications) is a more straightforward description or replication of the real world data form, with the qualitative leap of AI in the structure, understanding and accuracy of unstructured data, the fusion of structured and unstructured data by the so-called AI native application will be the norm in the future;

3. With the maturity of distributed technology, there is no major bottleneck in the storage and processing of "fact" data, but there will be a greater demand for secondary analysis and processing of different data, and even decision-making reasoning, and the demand for data volume, timeliness, accuracy, and intelligence will be higher and higher.

4. With the breakthrough and popularization of the bottleneck of distributed technology, the base of distributed databases will be gradually popularized and standardized like object storage/S3 in the future, and in application development, it will become possible to support various data structures on a set of storage, and the flexible use of application development will become easier. Developers no longer need to consider engineering complexities, including but not limited to operational complexity, reliability requirements, scaling bottlenecks, storage costs, transaction and consistency requirements, and so on.

We are also looking forward to the future world of AI Native, and we are constantly making breakthroughs and innovations to better meet the needs of these directions, and look forward to catching up with the general trend of the times and becoming the best data processing platform for modern application architecture. (This article was first published on the Titanium Media APP, author|Cai Pengcheng, Qin Conghui, editor|Liu Yangxue)