PolarDB for MySQL Enterprise Edition is a cloud-native Hybrid Transactional/Analytical Processing (HTAP) database built on decoupled compute and storage with shared distributed storage. It delivers high elasticity, petabyte-scale capacity, and sub-10-second failover while remaining 100% compatible with MySQL 5.6, 5.7, and 8.0.
How it works
A PolarDB cluster consists of one Primary Node that handles writes and up to 15 Read-only Nodes that handle reads, all sharing a single copy of data on distributed storage. This shared storage model means adding a Read-only Node requires no data replication — only compute resources are added.
The architecture has five key characteristics:
Decoupled compute and storage — Compute nodes store only metadata and process SQL requests. Data files and Redo Logs reside on remote distributed storage nodes, so compute and storage scale independently.
Shared distributed storage — All nodes in a cluster share one copy of data. Storage scales automatically based on actual data volume to petabyte scale, with no manual capacity planning.
Single-primary, multiple-replica architecture — One Primary Node handles writes; up to 15 Read-only Nodes handle reads. The built-in read/write splittingSQL mechanism routes SQL requests to the appropriate node automatically.
Data consistency and high availability — Storage nodes replicate data across multiple copies using the Parallel-Raft protocol for strong consistency. The Primary Node and Read-only Nodes use an active-active high availability mechanism. If the Primary Node fails, the system fails over to a Read-only Node in under 10 seconds — a Recovery Time Objective (RTO) of less than 10 seconds with zero data loss (Recovery Point Objective (RPO) = 0).
High-speed network interconnect — Compute and storage nodes communicate over a high-speed network using the Remote Direct Memory Access (RDMA) protocol, reducing I/O latency.
Core features
MySQL compatibility — Migrate existing MySQL 5.6, 5.7, and 8.0 databases without changing application code or configurations. PolarDB is also fully compatible with ApsaraDB RDS for MySQL. Beyond standard MySQL, it supports a single-writer/multi-reader or multi-writer/multi-reader architecture, shared storage, tables with up to 6 billion rows, DDL operations in seconds, seamless primary-replica switchover, and Flashback Query.
Native HTAP — Run transactions and real-time analytics on the same dataset in the same cluster. No Extract, Transform, Load (ETL) pipeline or separate analytical database is needed — data is available for analysis immediately after being written.
Multiple write nodes — In Multi-master Cluster (Limitless) Edition, all nodes accept read and write requests. Write throughput scales linearly as nodes are added. Node switchover takes 5–10 seconds without interrupting in-progress transactions.
Global active-active and geo-disaster recovery — Deploy clusters across regions for disaster recovery. During normal operation, the recovery cluster also serves read and write traffic to improve resource utilization. Data synchronization latency between clusters stays under 2 seconds, with minute-level failover for regional failures via the Global Database Network (GDN).
Lock-free Backup — Back up terabytes of data in minutes using distributed storage snapshots. The process requires no table locking and has minimal impact on running workloads.
Security — Protect data access, storage, and management with IP Allowlist, network isolation via Virtual Private Cloud (VPC), and SSL-encrypted connections.
Use cases
Handle traffic spikes without over-provisioning
For workloads with significant traffic peaks — such as e-commerce promotions or online course launches — PolarDB Serverless scales compute resources dynamically in seconds and adds or removes nodes in minutes. Storage is billed pay-as-you-go and scales automatically up to petabyte scale, so you provision for actual workload rather than peak capacity.
Maintain high availability for mission-critical systems
Finance, government, and enterprise ERP systems require continuous uptime and zero data loss. Configure a Cluster Edition cluster with at least one Read-only Node. With shared storage and multi-copy replication across availability zones, the cluster automatically fails over with RTO < 10 seconds and RPO = 0 if the Primary Node fails. For regional disasters, use GDN to achieve minute-level failover.
Modernize MySQL with no code changes
When existing MySQL systems hit performance or scaling limits, migrate to PolarDB. Full compatibility with MySQL 5.6, 5.7, and 8.0 means most applications migrate without code changes. After migration, workloads benefit from RDMA network throughput and distributed shared storage that scales to petabyte capacity. Lock-free Backup completes terabyte-scale backups in minutes without impacting online services.
Run real-time analytics on live transaction data
Enterprises that need real-time operational reports can use PolarDB's native HTAP capabilities rather than maintaining a separate "transactional database + analytical database" architecture. In the same cluster, the Primary Node handles transactions and a Read-only Node handles analytics. Data is available for analysis immediately after being written, with no ETL delay and no cross-workload performance interference.
Editions
PolarDB for MySQL Enterprise Edition offers two editions. See Editions for a full comparison.
| Edition | Write nodes | Best for |
|---|---|---|
| Cluster Edition | Single Primary Node | Teams that need high availability and horizontal read scaling for general-purpose or read-intensive workloads, with up to 15 Read-only Nodes |
| Multi-master Cluster | Multiple write nodes | Write-intensive workloads where write throughput must scale linearly as nodes are added |
Billing
PolarDB charges for compute nodes, storage space, and backup storage.
Compute costs
Billed by node specification and usage duration. Add or remove nodes and temporarily scale specifications up or down in minutes.
When adding a Read-only Node, pay only for the additional compute — storage is already shared.
Storage costs
PSL4 and PSL5 storage scale automatically based on actual data volume and are billed pay-as-you-go, with no upfront capacity planning.
PSL4 storage supports smart-SSD, which reduces storage costs by about 40% without affecting performance.
For full pricing details, see Product billing.
Get started
Manage PolarDB for MySQL clusters — including creating clusters, databases, and accounts — using any of the following:
Console — Graphical web interface for cluster management.
CLI — All console operations are available via the CLI.
SDK — All console operations are available via the SDK.
API — All console operations are available via the API.
After creating a cluster, connect to it using:
DMS — Connect to a PolarDB cluster using DMS to run queries in a web-based interface.
Client tools — Connect using standard MySQL-compatible clients such as Navicat or MySQL Workbench.