POLARDB - A database architecture for the cloud
In this session, we will give a introduction of PolarDB, which is a a cloud native database available on Alicloud. PolarDB separates the computing layer from a shared storage layer. This design allows the storage space to automatically scale according to usage. The computing layer adopts a a single master - multiple read only replica architecture, and scaling up a compute node or adding a replica only takes a few minutes. The computing nodes and storage nodes are connected with a high-speed network and use the RDMA protocol to execute data transmissions, resulting into I/O performance never being a bottleneck again.
A database in the cloud should be able to adapt both to increasing storage needs and variable load requirements. In this session we will discuss how POLARDB for Alibaba Cloud is designed to address these requirements. POLARDB separates the computing layer from a shared storage layer. This design allows the storage space to automatically scale according to usage. At the same time, the computing layer may be scaled independently using a single master and multiple read only replicas. Scaling up a compute node by adding a replica only takes a few minutes since no data needs to be copied. To avoid that the I/O performance becomes a bottleneck, the computing nodes and storage nodes are connected with a high-speed network and use the RDMA protocol to transmit data. We will also discuss how the storage engine for MySQL, InnoDB, has been adapted to using a shared storage by changing from logical replication to physical replication, and discuss the advantages of physical replication. Finally, we will present the ongoing project for supporting parallel query execution for POLARDB for MySQL. This will make it possible to efficiently run analytical queries.
Øystein Grøvlen is a Senior Staff Engineer in Alibaba Cloud, where he works in the Database Products Group.
Before joining Alibaba, Øystein worked for 10 years in the MySQL optimizer team at Sun/Oracle. At Sun Microsystems, he was also a contributor on the Apache Derby project and Sun’s Architectural Lead on Java DB. Prior to that, he worked for 10 years on development of Clustra, a highly available DBMS.
Øystein has been talking about the MySQL Optimizer at several conferences over the years. Dr. Grøvlen has a PhD in Computer Science from the Norwegian University of Science and Technology. Øystein is a native Norwegian, but currently lives in Hangzhou, China.