PolarDB:Editions

Cluster Edition is developed based on an architecture in which computing is decoupled from storage. It provides faster scaling, larger storage capacity, and lower replication latency.

• One primary node and multiple read-only nodes

  PolarDB for MySQL uses a distributed cluster-based architecture. A Cluster Edition cluster consists of one primary node and 0 to 15 read-only nodes. Multiple nodes constitute the engine layer of the database. The primary node processes read and write requests, and the read-only nodes process only read requests. Cluster Edition uses the active-active failover method between the primary node and read-only nodes. This method ensures high availability of databases.

• PolarStore

  Multiple compute nodes share one set of data, eliminating the need for each compute node to maintain a copy of the data. This significantly reduces storage costs. Cluster Edition uses a distributed storage and file system. This lets you seamlessly expand the storage capacity of your databases to hundreds of terabytes, without being restricted by the capabilities of individual database servers.

• Compute and storage decoupling

  Cluster Edition decouples computing from storage to let you easily scale clusters as your business grows. Compute nodes store only metadata, while data files and redo logs are stored in PolarStore. Only metadata that is related to redo logs needs to be synchronized among compute nodes. This reduces the replication latency between the primary node and read-only nodes. If the primary node is faulty, a read-only node can quickly take over as the primary node.

• Read/write splitting

  Cluster Edition comes with the read/write splitting feature that is implemented based on PolarProxy, so that it can perform transparent, high-availability, and adaptive load balancing. The read/write splitting feature automatically forwards SQL requests to each node of Cluster Edition based on cluster endpoints, allowing clusters to process large amounts of concurrent requests in an aggregated manner. For more information, see Read/write splitting.

• High-speed network connections

  Data is transmitted among compute nodes and storage nodes of a cluster by using the Remote Direct Memory Access (RDMA) protocol over high-speed networks. I/O is no longer a limiting factor to database performance.

• Data reliability and consistency

  The data on storage nodes has multiple replicas, which ensures data reliability. In addition, the Parallel-Raft protocol is used to ensure data consistency among these replicas.

The following figure shows the architecture of Cluster Edition.

The following figure shows the architecture of Multi-master Cluster (Database/Table) Edition.

Multi-master Cluster (Database/Table) Edition is suitable for high-concurrency read and write scenarios such as multi-tenancy, gaming, and e-commerce.

• Multitenancy in SaaS: high concurrency and load balancing between tenants

  Multi-master Cluster (Database/Table) Edition can implement load balancing by switching over databases of tenants across different primary nodes or adding new primary nodes to handle traffic bursts.

• Global gaming server and e-commerce scenarios: scaling in minutes to cater to fast-growing business

  Multi-master Cluster (Database/Table) Edition supports scale-out in seconds and provide the transparent routing feature. They can be combined with the middleware or business-based database and table sharding solution to shorten the scale-out process from days to minutes.

• Games deployed on different servers: better performance and scalability

  When a game is attracting large numbers of players, databases can be switched over to new primary nodes to implement load balancing. When the game is past its prime, databases can be aggregated to a few primary nodes to reduce operating costs.