ApsaraDB for RDS provides a variety of disaster tolerance solutions.

Multi-zone instances

Alibaba Cloud provides cloud computing services among multiple regions around the world. Each Region covers multiple zones. Zones in the same region have an extremely low network latency and are isolated upon failures.

Single-zone RDS instances run on two physical servers in the same zone, with redundant racks, air conditioners, circuits, and networks. Through asynchronous or semi-synchronous data replication and efficient master-slave switching, RDS provides users with database availability that is higher than the limits of physical servers.

ApsaraDB for RDS supports multi-zone to provide higher availability than single-zone. Physical servers are deployed across different zones. When a zone fails, traffic can be quickly switched to another zone. The entire switchover process is transparent, and does not require changes to the application code.

Note During a failover, the connection to all ApsaraDB for RDS instances is broken, and you need to reconnect the application to the ApsaraDB for RDS instance.

For information about migration zone, see Migrate an ApsaraDB RDS MySQL instance across zones.

Disaster recovery instance

The disaster recovery capability of a multi-zone RDS instance is limited to different zones in the same region. ApsaraDB for RDS also supports cross-region data disaster recovery. For example, you can asynchronously replicate instance A in the China (Hangzhou) region to instance B in the China (Shanghai) region by using Data Transmission service. Instance B is a complete and independent RDS instance, has independent connection addresses, accounts, and permissions.

After a disaster recovery instance is created, if the region where instance A is located encounters major fault that cannot be restored for short-term, you can switchover instance B in another region at any time. After the switchover is completed, you can change the database connection configurations in your application to redirect requests to instance B. This way, database availability above the region limit is guaranteed.

Note Before switchover, you must stop data replication from instance A to instance B to avoid data errors.

For more information about how to create a secondary instance, see Disaster recovery instances.