When ChatGPT was born, many industries were shocked and realized that the moment of change had arrived.
The database is one of the first areas to "wake up" and put it into action. It is generally believed in the industry that the powerful capabilities presented by large models make Data+AI the general trend in the database field. The resonance of large models and data at the same frequency will not only bring about a fundamental reconstruction of the current database technology architecture, but it is also expected to form differentiated competition in the database market and bring infinite possibilities.
Therefore, from the first half of this year, foreign database manufacturers such as DataBricks have been increasing Data+AI, while Chinese database companies have not given way. PingCAP, Alibaba Cloud, Huawei Cloud and many other companies have taken action to quickly respond to the trend of Data+AI .
As Liu Qi, founder and CEO of PingCAP, said at the recent PingCAP User Summit 2023: "AI is reshaping the software industry, and the database needs to be more systematically AI-ready in terms of architecture. When the database architecture achieves separation of calculations, storage When storage and storage are separated, it is very easy to introduce AI."
Liu Qi, Founder and CEO of PingCAP
Data+AI, Dongfeng of database differentiation
Kevin Kelly believes in "Twelve Future Trends" that we are now in an era of data flow. Business is the business of data. At the end of the day, all you are dealing with is data.
Indeed, as data becomes a new core factor of production, the R&D, supply chain, marketing, service, and even innovation of enterprises are all being reshaped based on data, while the continuous iteration of AIGC, large-scale models, and general artificial intelligence technologies is closely related to Evolution is expected to bring about a leap in productivity. The continuous integration of Data+AI will bring great changes to data and databases from the outside to the inside:
The first is that the threshold for data consumption has dropped at an unprecedented rate, and the use of data by everyone is no longer an unattainable vision. With the powerful capabilities demonstrated by large models and the great progress made in NLP to SQL and Text to SQL, an interactive revolution in data usage is taking place. Data usage and consumption are no longer exclusive to data scientists and other professionals. The consumption threshold has been greatly reduced.
And when everyone uses data to become a reality, it means that data integration, query, analysis, and application operations will show an exponential growth trend, which will undoubtedly bring about subversive changes in the architecture, performance, and reliability of database products. change.
Another significant change is that as data consumption becomes a new normal in the digital transformation of enterprises, the database itself needs to be service-oriented, shielding complex technologies and operations, and using AI to enable intelligent data management, use, and operations. and automation in response to new trends in data consumption.
"No matter how the technology world changes, stability, performance, high availability, ease of use, and tool ecology will always be important concerns of users for databases. We should move towards service-oriented databases, not service-oriented databases." PingCAP Founder and CTO Huang Dongxu said bluntly.
Huang Dongxu, founder and CTO of PingCAP
The author believes that the integration of Data+AI, while accelerating the reshaping of database products and architectures, Chinese database companies need to seize the opportunity to achieve all-round breakthroughs in products and markets. As we all know, due to the emergence of policy dividends, a large number of database-related companies have emerged in the Chinese market in the past few years. On the one hand, it shows the huge potential of the database market, and on the other hand, it also reflects the serious homogeneity of database products. At the time of technological change, it is not a long-term solution to enjoy policy dividends in a "partially safe corner". Only by achieving leadership in the core database architecture and taking root in the needs of market users can we truly walk out of the road of differentiation of database products, and then realize the market. A comprehensive breakthrough.
Facing the wave of Data+AI, PingCAP is the first to give the answer. At the PingCAP 2023 User Conference, PingCAP officially launched a database with a new generation architecture: TiDB Serverless. Liu Qi said bluntly: "TiDB Serverless started its pre-research four years ago, and after four years of polishing, it was officially released. At the architectural level, it meets the various new data processing needs brought by the AI era."
Database Accelerates towards AI Ready
How does the database accelerate towards AI Ready? It can be said that cloud native is an unavoidable topic.
As we all know, with the widespread popularity of cloud computing, a large number of traditional government and enterprise users have begun to use the cloud, and driven by the application of AI technology, an era of comprehensive cloud nativeization has arrived: from infrastructure to applications, it has begun to move towards Cloud native. Among them, the database is another important starting point. The continuous evolution of the database from the monomer to the microservice architecture to the serverless architecture responds to the changes in the requirements of comprehensive cloud nativeization.
If cloud native is the basic condition for databases to become intelligent, then AI is the driving force for the continuous evolution of cloud native databases. Just as new energy + intelligence is reshaping the automotive industry, cloud native and AI are also profoundly affecting database architecture and products.
Taking the TiDB Serverless database as an example, it realizes the automatic processing of complex transactions and greatly simplifies the development of applications. Users no longer need to spend a lot of time on database processing, so they can devote their energy to business innovation.
It is reported that TiDB Serverless adopts the cloud-native/multi-cloud design concept, and has a cloud-native engine CSE (Cloud-native Storage Engine) architecture, which can realize database services that do not require resource planning, start in seconds, start at 0 yuan, pay per use, and are extremely elastic The key components of TiDB Serverless adopt a fully separated design, which not only has automatic resource scheduling capabilities, but also can flexibly integrate AI capabilities; in addition, new functions such as Chat2Query have greatly reduced the threshold for data consumption, forming a good relationship between the database and the data consumer. butt.
"TiDB Serverless is suitable for multiple application scenarios. It is extremely convenient to use and can achieve efficient cost control." Liu Qi said: "In just a few months since its launch at the end of last year, the TiDB Serverless beta version has more than 10,000 active users. Cluster."
Gartner predicts that by 2025, the proportion of cloud-native load will reach 95%, and the proportion of new cloud-native applications will continue to increase in the next few years. The author believes that in a fully cloud-native era and driven by the demand for AI applications, traditional database vendors will gradually lose their original advantages, and databases born in the cloud, longer than the cloud, and based on cloud-native architecture will become the backbone of the market . "Databases are fully integrating trends such as Serverless and AI. TiDB Serverless can be integrated with HTAP and AI to form a trinity of innovative advantages." Liu Song, vice president of PingCAP, said.
But that doesn't mean anyone can roll out a true cloud-native database anytime soon. As we all know, databases have long been a field with a high degree of engineering. Cloud-native databases not only need to be forward-looking in the selection of initial technical routes, but also need to continuously and repeatedly polish products. There is no doubt that PingCAP is at the forefront of the industry in terms of the advancement of database architecture and the integration of AI technology, and has pioneered a leading path for Chinese database products.
Liu Qi believes that there is still huge room for exploration in the combination of database and AI, such as workload prediction, intelligent resource allocation, data analysis, etc., "Small models may play great value in many aspects of the database. Small models are now making great progress. Faster, more specialized and compliant, especially the continuous improvement of small model reasoning capabilities, will greatly promote the usability of AI in the database field."
Take root in China and realize the two-way rush of business and technical value
According to IDC, in 2022, the scale of the database generated in the Chinese market will reach 23.3ZB, accounting for 23% of the world, and it is expected to become the world's largest data circle in 2026. With the steady advancement of the construction of digital China, the accelerated development of the digital economy and industrial digitization will continue to bring about explosive growth in data scale and rapid deepening of data applications, which will inevitably pose higher challenges to database innovation.
In the author's opinion, the next five years will be a critical period for China's database industry to break through. With the continuous deepening of the digitalization process of thousands of industries in China, the business scale, business complexity and market demand of the Chinese market will be unmatched in the world. For example, the Apps of many financial institutions such as banks in China have reached the scale of 100 million monthly active users (MAU), and a large number of financial institutions' marketing, customer service, risk control and other businesses are based on data and AI technology to achieve... Various changes, After the Chinese database market has experienced the influence of factors driven by macro policies, the future market will truly be driven by advanced technology.
Undoubtedly, PingCAP's adjustments to its market and product strategies this year are expected to allow it to better cultivate the Chinese market and use database products with advanced technology and architecture to help the digital transformation of thousands of industries.
First of all, PingCAP has formed a clearer market and product strategy. Through the three major products of TiDB Enterprise Edition, TiDB Cloud, and TiDB Serverless, as well as the Pingkai database for the Chinese market, it can meet the needs of different markets and user groups for database use. The style of play in the market is more clear and focused.
Secondly, multiple product lines of PingCAP are based on a core core of TiDB. Although different products have different market positioning, they all maintain the advanced nature of the core. Taking the Pingkai database for Chinese enterprise users as an example, on the basis of TiDB Open Core, a large number of enterprise-level features such as localized basic suites and enterprise-level security suites have been added to meet the differentiated needs of the Chinese enterprise-level market .
Finally, PingCAP continues to improve in terms of localization ecology, delivery and service systemization. Pingcap Database has invested a lot of energy in localization ecology compatibility. While meeting the needs of Chinese enterprises for digital transformation, it also ensures that advanced database products can Get full play in the local technology ecosystem.
"Pingkai database was born out of TiDB Enterprise Edition. TiDB Enterprise Edition has been polished for five years. In the past, it provided more general functions for global users, but this is far from enough for Chinese enterprise users. The launch of Pingkai Database can provide Chinese enterprise users bring long-term value." Liu Qi finally said.