In the past, no matter the operator or the core system of the bank, the big architecture was monopolized by three big mountains of western IOE(IBM, Oracle, EMC). Until 2008, Ali put forward the "IOE" movement and began to boost the wave of localization of information software.
Tian Yun data was one of the earliest entrants. In 20 10, in order to establish a complete cloud computing industry chain in China, Tian Suning, the father of broadband in China, invested in building a cloud base, from which Tian Yun data hatched and began to take shape.
In 20 15, Lei Tao led the founding team to formally establish Tian Yun Data, taking the lead in cutting into the financial field. Tian Yun provided Hubble, the leading HTAP database in China, which completed the most difficult part of "IOE removal", replaced the traditional western IOE architecture of financial A-class core system, and solved the problem of reducing the burden of A-class core system in online banking transactions. In addition, in order to lower the threshold of using AI, Tian Yun Data also launched MaximAI, an AI PaaS platform, which gradually expanded the data value to other industries such as energy, medicine and military.
At present, Tian Yun Data has more than 70 large enterprise customers in the industry, with a single contract of 2-5 million, and the annual revenue of pure software exceeds 100 million.
In terms of financing, Tian Yun Data received an investment of 65,438,000,000,000 yuan from Xiyu Capital and Huaying Capital B in 2065,438+08.
As an industry veteran, Lei Tao has more than 20 years of technical management experience in North American multinational companies. In 2005, he joined the China Technical Committee of SNIA Storage Industry Association and was a member of the Big Data Committee of CCF China Computer Federation.
During the period of 20 1 1 cloud base, Lei Tao and the founding team were responsible for the business of many operators through the BDP big data platform, such as Unicom's data cube, mobile headquarters and southern base. 20 15 After Tian Yun Data became independent, Lei Tao chose to focus on the financial field first in order to avoid horizontal competition.
The goal of Tian Yun Data is to replace Oracle Bone Inscriptions and SAS. The accumulation of cloud-based period made Tian Yun data have a high starting point from the beginning, and the first order took over the ——OLTP special line trading system, the core system of China Everbright Bank. For example, banks can realize OOTD transactions in all business halls across the country in real time, and query the amount of deposits and withdrawals in real time. The technologies involved in the whole process are some substitutes for Oracle in the early days of Tian Yun Data.
However, during the operation of many projects, Lei Tao found that under the strong consistency of millions of transaction specifications, the mobility of data, the change of computing framework and online transactions all require large-scale parallel computing at the same time, which requires high universality, immediacy and total data volume of computing scenarios, and the traditional Oracle architecture can not adapt at all.
On top of Oracle architecture, it needs to be upgraded to meet the new requirements.
So Tian Yun Data independently developed the domestic distributed database Hubble of HTAP. Different from the traditional IT architecture, which needs online analysis and separate processing, HTAP database can support business systems to run and do OLAP scenarios on the same data at the same time, avoiding a lot of data interaction between online and offline databases and reducing the burden on the system.
Hubble, a distributed database made by HTAP, has replaced the all-in-one Oracle, with more than 2,000 80T core tables, 40 billion transaction data, 56 service application transactions, 500 concurrent users and 500ms transaction service responses. The daily online transaction volume exceeds 2 million transactions, accounting for 65,438+00% of the bank's core transaction volume, enabling the bank to provide 7*8 hours of Class A real-time core transactions for the counter system.
Converting from centralized Oracle to distributed HTAP also solves the problem of database scalability. For example, Tian Yun data enabled China Everbright Bank to solve the problem of historical data query. In the past, only the historical query of two years ago could be found, but after the distributed technology went online, all transaction data before 15 could be queried, and at the same time, the bank counter system and mobile APP could also be queried by countless people at the same time.
In the process of BI gradually turning to AI, complex business processes are reconstructed by algorithms. In the past, it was necessary to take the data to SAS platform for analysis and put it layer by layer. But now through distributed technology, the process tends to be flat, and millisecond service response can be realized.
Tian Yun data has shaken the industry's head resources from the very beginning. At present, Tian Yun Data has large enterprise customers in more than 70 industries, such as China Everbright Bank, Industrial Bank, China CITIC Bank, Zhongtai Securities, China Petroleum and National Bureau of Statistics, which are distributed in the fields of finance, energy, medicine, government and military affairs, with a single contract level exceeding one million.
For each vertical industry, Tian Yun Data will set up a subsidiary to focus on the track. At present, Tian Yunyou has 160 people, and technicians account for more than 60%.
In Lei Tao's view, if 600 projects a year are fragmented orders of 50,000 yuan and10.5 million yuan, the company always meets the simple needs of primary customers repeatedly, and the technology is difficult to precipitate and deepen. At the current growth stage, building products needs to find a balance between what users want and what you want to do.
For Lei Tao, focusing on the development of head big B has two great development potentials. On the one hand, Big B has the general ability and laboratory of machine learning, and it is easier to accept new products. On the other hand, while delivering products and services, Tian Yun Data is also transferring the data value of Big B customers.
"AI itself is a knowledge production process. It can quickly sample and copy the experience values of large enterprise rules and processes, and empower other customers in the industry and even other similar industries. "
However, with the head customers becoming more customized and personalized, has Tian Yun data lost its powerful replication ability?
Lei Tao explained that although the requirements of each enterprise are different, they are all looking for a database in a small pool. Enterprises can migrate, clean and copy data from massive data, and can find suitable AI methods to make it produce commercial value. This process is universal.
Speaking of core barriers, Lei Tao thinks that the data barrier in Tian Yun is the replication value of data.
The construction of the barrier can be divided into two stages. The first stage is the barrier of cutting-edge technology itself, which compares efficiency and core value of products. Whoever can go deep and deliver better will win the first prize. As the earliest research and development team of big data and artificial intelligence in China, Tian Yun Data has a certain technological first-Mover advantage.
The second stage is the service of reasoning end. The value of data resources needs to be refined through machine learning to form knowledge, and then packaged into reasoning services to serve the industry. For example, the characteristics and contents learned from the 20-year critical illness compensation pricing of an insurance company can be quickly transplanted to the insurance industry, and large enterprise customers bring high-quality training databases to Tian Yun Data.
In the future, AI will detonate the trillion-dollar market, but the current penetration rate is less than 1%, which leaves many opportunities and imagination for enterprises. But no matter what kind of enclosure, the final comparison is speed, service stability and product capability.