(图片来源:unsplash)
Recently, Robin Li, the founder, chairman and CEO of Baidu, said that the open source model will become more and more backward, causing widespread discussion in the industry about 'open source' technology. Not only in the field of AI large models, but also important practitioners in many fields, including open-source databases, have also expressed their views on this.
At the OceanBase Developer Conference in Shanghai on April 20, Yang Chuanhui, Chief Technology Officer (CTO) of OceanBase, told Titanium Media App that from the perspective of the database field alone, open source technology is very important to the industry, and both open source and commercial databases have a certain space for survival.
"As far as open source and commercialization positioning are concerned, I think that first of all, I must do a good job of open source, and I must make the things out of open source sincere enough to truly solve the core pain points of users; Here in OB, the database is equivalent to an invention of a bicycle, 'open source' is to teach you how to ride a bicycle, 'commercial' is to sell you a good bicycle, but you can't ride useless, open source is helpful to our commercialization. There's no problem with OB, we can mix open source and business well. Yang Chuanhui said.
Yang Chuanhui, Chief Technology Officer (CTO) of OceanBase
In fact, as a 100% self-developed distributed database, OceanBase continues to implement the product strategy of integrated database to load business-critical systems.
In 2020, OceanBase announced its independent commercial operation. At that time, OceanBase said that by adding analytical business processing capabilities to the original functions, OceanBase would be upgraded to an enterprise-level distributed database that supports HTAP mixed workloads.
In 2022, OceanBase 4.0 was released, breaking through the performance bottleneck of a distributed database for the first time and implementing the distributed integration of a single server. In terms of integration, OceanBase has implemented single-node distributed integration, TP/AP integration, cloud-on-cloud and on-premise integration, and multi-mode integration, using a single database to meet the needs of customers in 80% of scenarios.
In April 2024, OceanBase 4.3 is released, and a columnar storage engine is released to build a petabyte-level real-time analysis database, which can achieve real-time analysis in seconds and further strengthen the integration of TP and AP. The on-site score shows that under the same hardware conditions, the AP query performance of OceanBase 4.3 has reached the same level as the industry's first-class column-stored large-width table database.
At the OceanBase developer conference, Yang Chuanhui said that OceanBase 4.3's further breakthrough in TP/AP integration is also the result of OceanBase's insistence on 100% self-developed products. OceanBase has complete control over the kernel code and has a stronger ability to break through complex business scenarios and support key business systems by not building a secondary development environment and process based on open source databases. This marks an important step forward in its "mission-critical" integration strategy.
It is worth mentioning that in March this year, Ant Group announced that OceanBase has established a board of directors to face the market independently.
Talking about this, Yang Chuanhui said, "This shows that Ant is more determined and more invested in the long-term sustainability of a database, which is a core." Based on this core, I think it is good for our product technology. Today, you can see that OceanBase has been recognized in the industry, but it has not yet been completely successful. We believe that the integrated architecture has the opportunity to become the new mainstream, both at home and abroad, and on a global scale. Based on such thinking and judgment, we have the confidence to make it operate independently and accept the test of the market. ”
"In order to really do a good job in the database, we must first dare to challenge the most difficult things and dare to develop our own TP (transaction processing, transaction-oriented processing system). TP barriers are extremely strong, and the accumulation of TP barriers and user scenarios will make a database form a particularly large ecosystem. In this process, we are very competitive and do not want to be 'one of a kind'. From a business point of view, some companies decline because there is no competition, not because there is a lot of competition, and healthy competition is good for all participants. Yang Chuanhui said.
At present, the OceanBase Landscape technology ecosystem has expanded to more than 750 mainstream products, covering infrastructure, data integration, data governance, application integration, and other fields. OceanBase has served more than 1,000 industry customers, 30% of which are applied to core systems.
Talking about the new round of AI craze, Yang Chuanhui pointed out that the GPT large model has a great impact on the database, although OceanBase does not specifically support vector database products, but the entire plug-in and community has the ability to vector database, and the open source community has tried to do vector demo based on OB open source. "In the future, we will go in this direction to lay out research and development, and now there are some attempts. ”
Chen Wenguang, a professor at Tsinghua University and dean of the Ant Technology Research Institute, said that the development and rise of AI large models have also put forward many new requirements for the entire data processing and database fields.
Looking to the future, Chen Wenguang said in his speech that there are three new trends in the field of databases in the future: online and offline integration, the integration of vector databases and relational databases, and the greater vision is the integration of data processing and AI computing.
"In the past, data processing was more done in a low-level way, and then it was collected and stored first, and then bookkeeping, and then some analysis and aggregation were done on this data. AI has taken up more and more of the cost and capacity of the entire computing, if this is the case, what should be the relationship between our data processing and AI? I think it should also have a trend of integration. Chen Wenguang said.
At the end of his speech, Yang Chuanhui said, "I hope that in the future, there will be more and more interesting attempts like 'vector database' in the OceanBase open source community, and our open source community is not a cold product, but a real community that connects you and me, and allows everyone to make friends and have fun." ”
(This article was first published in Titanium Media App, author|Lin Zhijia)