Ant architect Guo Yuanfei: distributed database is the best way for financial institutions to digital transformation


Introduction:Oceanbase is determined to become the world’s leading provider of enterprise data technology solutions.

On August 26-28, 2020, Guo Yuanfei, senior solution architect of ant group, delivered a keynote speech entitled “oceanbase distributed relational database helps insurance business innovation” at the China Property Insurance Technology Application summit held by China Science and Technology Co., Ltd., sharing the development history, product value and insurance of oceanbase, an enterprise level distributed relational database of ant group Industry best practice applications. The following is a summary of the speech.

Ant architect Guo Yuanfei: distributed database is the best way for financial institutions to digital transformation

Good afternoon, dear friends!

Alipay has more than a billion users worldwide, and they can simultaneously pay for transactions reliably, stably and efficiently. Behind this cannot do without an important secret weapon, OceanBase. Today, I am honored to introduce the development process, product features and best practices of oceanbase in the insurance industry.

Iteration of distributed database oceanbase

Oceanbase database was established in 2010, and the first customer was Taobao favorites. In 2013, oceanbase started support for SQL. In 2014, oceanbase began to support all business of e-commerce bank, and became the first distributed relational database applied to the financial core business system in the world. From 2014 to 2016, Alipay’s entire trading system and core accounting system ran on top of OceanBase.Over the years, all the transactions of the world’s hundreds of millions of users on Alipay are carried by OceanBase and remain rock solid.

In 2017, we began to export our technology to more external customers, and we helped Bank of Nanjing build the core of “Xinyun +” Internet financial business. These scenarios that oceanbase has served include not only the emerging Internet financial core business, traditional financial core business, but also the replacement of traditional database and MYSQL, which are very successful.

In 2019, oceanbase participated in the most authoritative TPC-C certification test in the world database, and we achieved 60.88 million TPMC, twice as much as Oracle, the second traditional database company. TPC-C is absolutely not a simple run score test. In fact, it is a test that requires the tested object to meet the four transaction characteristics of acid database.

In 2020, we once again took part in the TPC-C test, reaching 707 million TPMC, running 12 times as much as last year, and once again won the top of the TPC-C list. In June this year, oceanbase officially became an independent company, and we are determined to become the world’s leading provider of enterprise level data technology solutions.

Database morphology evolution

The core business system of financial industry, including the insurance industry, still uses traditional centralized database on a large scale. This scheme brings two problems, one isThe cost of centralized database is very highOn the one hand, software license is very expensive; on the other hand, hardware is very expensive, because it must use high-end hardware, such as IBM mainframe and EMC high-end storage, which are very expensive, and it is difficult for enterprises with a long time to bear; second, it is very expensive for enterprises with a long time to bear; second, it is very expensive for enterprises with a long time to bear; second, it is very expensive to use high-end hardwareVery poor scalabilityTraditional centralized database is based on shared storage to do horizontal expansion. When facing the development of key business systems, this scheme is powerless and difficult to provide effective support.

For the above solution, the industry has proposed a patch solution. More than a decade ago, it was proposed that since the traditional centralized database is very expensive, then add a syntax compatibility on the basis of stand-alone open source to solve its expensive problem. But today, we have not seen the widespread use of this scheme in the core system of the financial field. The reason is very simple. The core system of the financial industry uses traditional centralized databases not because of their perfect syntax, but because they are stable and reliable. This patch solution can not do this, and this solution does not solve the problem of scalability.

When it comes to scalability, it was also more than ten years ago that such a scheme was first put forward in Internet enterprises, which is called sub database and sub table. A typical example is to integrate MySQL single machine into MySQL Cluster with proxy, which provides good computing power. This solution also has great problems. When the calculation takes place in a single MySQL machine, everything is fine, but once the user’s load crosses the boundary of a single MySQL machine, you may find that the performance of the sub database and sub table system is not satisfactory. Specifically, it can not support global transactions, strong data consistency, load balancing and complex SQL, so the usage scenarios of this scheme are very limited. Up to today, this scheme has not been widely used in the core business systems of banking, insurance and securities companies.

So is there such a database system? It can not only provide high reliability of traditional centralized database, but also has controllable cost and good scalability. On this basis, it can also meet the requirements of transaction acid, strong consistency and complex SQL processing ability.The answer is yes, and it can be better, that is oceanbase.

Three unique values of oceanbase

First, oceanbase is a distributed relational database that is fully independently developed,The code is 100% self researched, including SQL Engine and storage engine, which are written by our engineers one line by one. This is important. It means that when your database is in trouble, our engineers don’t need to turn to the open source community, and we can go to the bottom of it. This is a new requirement for the core business system in the financial field that requires stability, reliability and predictability. Oceanbase can provide it fully.

Second, oceanbase is the only native distributed database in the world that has been verified by long-term complex financial scenarios.In fact, database is a very complex system, and its complexity is no less than that of the operating system. The testing of its various complex application scenarios is not a simple test program that can be exhausted. At present, the only effective way is to find enough complex and enough scenarios in practice to repeatedly use the database, until there is no problem in using it, we will trust it. We are very happy to tell you that oceanbase has done it.

Third, oceanbase database is the only native distributed database in the world based on the ordinary x86 cloud server, which ranks first in the TPC-C field.Topping the TPC-C list means that oceanbase can not only provide world-class performance, but also provide a complete experience as a stand-alone database, which means that users can use oceanbase just like a stand-alone database.

Core features of oceanbase

In oceanbase database, we define the leading standard of distributed database through the characteristics of “five high and one more”.

1. High availability
Through Paxos protocol and multi replica technology, we can provide good high availability. When there is a disaster, we can achieve RPO = 0, RTO < 30 seconds, which can fully meet the disaster recovery needs of enterprises. At the same time, because we use Paxos protocol, Paxos protocol supports log hole, which brings a benefit. When the network fluctuates, we can guarantee better performance.

The high availability of oceanbase is completely provided on the general x86 server and ordinary hardware, without using any high-end hardware, so our cost will be lower.

2. High expansion
In the actual production system, the servers in a single oceanbase cluster can reach more than 100. And oceanbase has a good linear expansion ratio. High scalability is also reflected in the automatic load balancing during capacity expansion, which means that when DBA needs to expand, all we do is add a machine to the cluster. Next, our oceanbase management system will automatically migrate data to the new machine according to the load situation of the system. The whole process of capacity expansion and capacity reduction, the business has no perception.

3. High performance
We use a lot of advanced technology to improve the performance of the database. For example, LSM tree, lock free architecture, eliminating random writes to disks, and so on. These technologies help us make full use of the capabilities of our hardware, coupled with high scalability, we can provide a world-class performance oceanbase cluster.

In the actual production system, we can provide 61million times per second at peak time, and the maximum capacity of single table can reach 320billion rows. With high performance, it is low cost. Because we adopt LSM tree structure, it is more organized when data is dropped, and can achieve 1:3 compression ratio.

4. High transparency
We have implemented many key technologies, such as global consistent snapshot, global index and automatic transaction two-phase commit. With oceanbase database, the application is just like using a stand-alone database, without special perception and modification for distributed database.

5. High compatibility
On one set of oceanbase cluster, we provide you with two sets of ecology, one is Oracle ecology, the other is MySQL ecology, which effectively reduces the cost of business migration and transformation. At the same time, we have done mutual recognition support with domestic mainstream operating systems and chips, which can effectively meet the demand of technology supply chain security.

6. Multi tenant
Multi tenant means that many tenants can be established on an oceanbase cluster, and each tenant corresponds to an Oracle instance or a MySQL instance. This feature can be used to achieve large-scale business centralization, and centralize the traditional centralized databases in one oceanbase database cluster. It can also centralize different kinds of small businesses. The previous dozens of small businesses are now centralized in an oceanbase cluster for unified management. This feature can not only effectively utilize the cluster resources, but also effectively reduce the operation and maintenance costs of DBAs.

Product orientation and form of oceanbase

Oceanbase can provide an OLTP engine with world-class performance, as well as an excellent OLAP engine. This means that with the multi tenant function mentioned just now, you can create different instances in an oceanbase cluster, some of which are dedicated to transactions, some are dedicated to analysis, and of course, you can also create a large instance for both transaction and analysis. At present, oceanbase has really implemented a set of HTAP engine, which can support both transaction and analytical business.

The disaster recovery architecture of oceanbase supports both the classic primary and standby disaster recovery methods and the disaster recovery methods based on Paxos protocol. We support three copies in the same computer room, three computer rooms in the same city, or even three centers in two places, three places and five centers. These different disaster recovery levels can fully meet your disaster recovery demands.

As for the deployment mode, if you want to use it with other excellent products of Alibaba cloud, such as sofa, a financial level distributed middleware, or OSS for object cloud storage, then oceanbase can be exported to you on Alibaba cloud’s proprietary cloud. If you only want to use oceanbase, we can provide it to you in the form of a pure license. We can install our software on your server, and you can get a high availability, high expansion and high performance cluster. If you want to purchase only one public cloud service, you can purchase an RDS of oceanbase on the public cloud of alicloud. These different options can seamlessly adapt to your different needs.

Practice of oceanbase in insurance industry

Oceanbase’s best practices in the insurance industry. A good case is human health insurance. If an unfamiliar friend can open your mobile phone Alipay, you can see the health care insurance products on the above health insurance. PICC Health uses the financial level distributed database oceanbase and financial level distributed middleware sofastack to solve the problem of slow processing capacity of high concurrency insurance policies. At present, the average daily order volume is about 15W.

Since the successful launch of people’s health care insurance and health insurance products in April 11, 2018, good health insurance has become Alipay’s insurance product. After oceanbase was launched, PICC has made significant benefits

  • The processing capacity has been increased by thousands of times, and the capacity can be dynamically expanded according to the business situation
  • 2018 medical insurance in-patient medical insurance, daily settlement documents of 10000 documents a day, the processing speed is reduced from the previous 4 hours to 6 minutes
  • The launch time of new products is shortened by more than 80%, from weeks to days
  • Using a variety of protocol conversion and standardized docking process, the external channel access efficiency is increased by 6 times, and the third-party business is rapidly expanded

Thank you for listening to my report and looking forward to using oceanbase database to serve the insurance industry better in the future, and build a core business system for the future insurance industry with all good friends.

Link to original text
This article is the original content of Alibaba cloud and can not be reproduced without permission.