17 solstice December 18, 2020, “Distributed Cloud | 2020 global Distributed Cloud conference” officially kicked off in shenzhen. The conference discussed several popular concepts such as “cloud native”, “distributed cloud”, “distributed data” and “distributed storage”. On the morning of December 17th, at the “Distributed Cloud Main Forum”, Dongxin Wu, senior solution architect of OceanBase Solutions Department of Ant Group, gave a speech titled “Native distributed Database brings Data Management Revolution” to the audience.
01 Accumulate experience with the pace of The Times
With the digital transformation, cloud computing, edge computing and the rise of various Internet businesses, a large number of businesses need to use database, which poses a huge challenge to the processing capacity of database. Traditional IOE database relies on high-end hardware, which makes the system difficult to expand and expensive. Although the solution of database and table based on middleware solves the problem of expansibility, cross-library transactions, global consistency, load balancing and complex SQL still become the problem to be solved urgently.
Dongxin Wu, Senior solution Architect, OceanBase
The emergence of distributed database solves the problem of horizontal linear expansion of database processing capability. Distributed database based on common server, reduce system cost, and support native distributed query. Transparently supports distributed transactions for applications, ensures global consistency, and has flexible deployment modes and load balancing capabilities.
Under the trend of rapid development of distributed database, Ant Group quickens the pace of development. Since the project was officially approved in 2010, Taobao Favorites became the first user of OceanBase, which was subsequently applied in the core system of Ant Group on a large scale, creating a record of 61 million TPMCS in TPC-C in 2019. In June this year, OceanBase formally set up an independent company, Beijing Aoxing Bess Technology Co., LTD., to carry out independent commercial operation. “Ant wants the native distributed database capability to benefit more customers in more scenarios, so it started the journey of commercializing OceanBase distributed database to serve external customers,” said Wu Dongxing. Today, we can scale fully distributed, multi-site.”
For OceanBase, from the internal use on the first day to the scene of supporting a large number of transactions with money on Double 11, it gradually solved the expansion capacity, disaster recovery capacity, and finally solved the large-scale operation and maintenance efficiency. This is the difference between OceanBase and other similar products. OceanBase was incubated by Alipay, and experienced the polishing of large-scale extreme scenes inside, and then fully opened to the outside after the product gradually matured.
02 Multi-form technology layout to achieve double drive market technology
By 2020, OceanBase will pay more attention to market drive. Aiming at customers with relevant demands of medium and large domestic financial institutions, governments and public utilities related to national economy and people’s livelihood, OceanBase can solve customers’ demands for high availability and strong consistency. Meanwhile, OceanBase can realize the transparency of upper-layer applications and reduce the intrusion of applications. It can also expand automatic sharding, online capacity expansion/reduction and other functions. Wu Dongxin said, “In terms of the technology development of the native distributed database, we hope to go cloud native, and we will also do storage and computing separation in the future.”
OceanBase is 100% proprietary database software. Wu Dongxin introduced, from the entire storage engine to the transparent distributed multi-site multi-center multi-active deployment architecture to the system engine are all self-developed. Although there are two application modes of Oracle compatibility and MySQL compatibility, OceanBase achieved tPC-C 707 million tpmC world record by horizontal expansion through native distributed architecture.
OceanBase database architecture has many characteristics, such as multiple copies, peer nodes, no storage device sharing, partition-level high availability, and strong consistency. In the implementation of multi-copy architecture, all nodes are peer, collaborative services can be fully read and write, and do not need dedicated devices and dedicated storage network, can realize the way of multi-site and multi-live. Wu Dongxin said that although OceanBase is a distributed database, users will find that the usage mode of OceanBase is exactly the same as that of traditional stand-alone database, without special modification and special adaptation by applications. The difficulty is handled by the database and the convenience is left to applications.
OceanBase is far superior to traditional architecture in terms of availability. If a few copies or a ZONE fails, the writing transaction succeeds when more than half of the copies are successful and the services are not affected. If the master copy is abnormal, the remaining majority of slave copies will negotiate a new master copy according to the Paxos protocol. “In the multi-site, multi-site mode, zero data loss can be guaranteed as long as more than half of the replicas survive, whether it is a machine room failure or a multi-site failure,” Wu said. Even if the whole machine room goes down, the interruption of optical fiber and network can achieve RPO=0 within 30 seconds.
We provide a very rich here live solutions, including if only one room, we provide high availability is a room, into the center of dual three traditional deployment architecture, live to other cities, including other customers can’t provide more room, we also provide main case library architecture of city double rooms, and provide all kinds of disaster and protection of the business continuity.
OceanBase has several typical deployment modes. For example, if three equipment rooms are deployed in one city, OceanBase can implement load balancing and disaster recovery (Dr) protection across equipment rooms. The latency of multiple core equipment rooms in the same city is generally 0.5 to 2ms. Normally, the delay of the two-site three-center deployment is the same as that of the three-center deployment in the same city. The two-site, three-center, and five-copy deployment does not affect the service response delay when a single ZONE fault occurs. OceanBase takes the distributed multi-activity architecture of three places and five centers as the ultimate pursuit of “no data loss, no business shutdown”. “This year, we added the active-standby database solution to provide customers with available DISASTER recovery solutions at a relatively low cost.” Wu Dongxin said.
At the end of the speech, Wu Dongxin concluded: “OceanBase, as an original and original distributed database, our biggest hope is to let everyone enjoy the benefits of distributed database, and at the same time, the application development is as transparent as the original stand-alone database, and there is no need to do a lot of application adaptation. We use innovative partition-level live to ensure that all resources do not have a read-only model, ensure that all hardware resources are fully utilized, and provide full transparency across multiple sites live.”
To sum up, OceanBase’s partition-based data protection approach really allows distributed databases to be adjusted on demand, making online capacity expansion and reduction a reality for the future data management revolution of native distributed databases, said Wu dongxin. In the future, OceanBase also hopes to work with more industry partners with vision and ability to build ecology together.