OceanBase: A New Distributed RDBMS behind AntFinancial's TechFin Technology
AntFinancial OceanBase Team
14 Mar 2018 Wednesday, 02:00 PM to 03:00 PM
COM2 Level 4
Executive Classroom, COM2-04-02
Alibaba's 11.11 shopping festival has become a landmark event in the development of China e-commerce. Behind the ever-growing gross merchandise volume (a.k.a GMV) is the rapid development of AntFinancial's TechFin Technology. Starting from 2014, OceanBase - a new distributed RDBMS developed by Alibaba & AntFinancial - has become an integral part of AntFinancial's technology infrastructure. Over the years, the entire AntFinancials's core trading system has been migrated from the support of commercial databases to OceanBase. In the last 11.11 shopping festival, the system made a world record of 256000 transactions per second as the peak throughput, which equals to 42 million SQL requests per second. In 2017, OceanBase became available to users outside AntFinancials. Customers, like Nanjing Bank, China Zheshang Bank and etc., started to build their own internet financial systems based on OceanBase.
OceanBase is a share-nothing distributed RDBMS built from the ground up since 2010 by Alibaba & AntFinancial. Rather than relying on high-end storage like traditional commercial databases, OceanBase use commodity PC's to offer superior scalability and minimize the hardware cost. Data consistency is guaranteed across geographically dispersed IDC's using PAXOS protocol to sustain failures from single server, data center or even the entire city. The design of the storage layer follows the classic 'LSM-tree' model, where data modification are kept in memory and 'merged' back to baseline data periodically. This design generates an ideal workload for SSD in that all random writes are eliminated and write amplification is kept to its minimum. Besides the linear scalability, superior performance and high availability, OceanBase also features cost-based optimizer, distributed query execution and other important capabilities to support hybrid OLTP&OLAP workload. This talk will give a global picture of the system and introduce some of the technology challenges that we faced over the years and how we managed to solve them.
Mengmeng Chen is the technical leader of the OceanBase SQL team. He received his B.Sc. degree from Tsinghua University and M.Sc. degree in Computer Science from University of Wisconsin-Madison. He first worked for Oracle in the Parallel Query group, focusing on the optimization and execution of parallel query. In 2012, he joined Huawei and worked on a research project on columnar databases. In 2014, he joined OceanBase team and led the design and development of the SQL execution engine. Currently, his team's responsibilities include query processing and optimization, distributed query processing, performance optimization and etc.