当您根据上一节中的表定义了组织的灾难恢复战略后，请选择最适合您的RPO和RTO要求的数据库。默认情况下，所有AWS数据库服务都提供备份和恢复功能。使用下表了解AWS数据库服务提供的更高级灾难恢复功能，并将其与您的灾难恢复需求相匹配。
| AWS数据库服务 | 到灾难恢复区域的复制方法 | 可能的备用分类 | RTO | RPO |
|---|---|---|---|---|
| Amazon RDS for MySQL | Cross-Region read replica |
|
Usually minutes. Automation can minimize delays. | Typically under 1 second |
| Amazon RDS for PostgreSQL | Cross-Region read replica |
|
Usually minutes. Automation can minimize delays. | Typically under 1 second |
| Amazon RDS for MariaDB | Cross-Region read replica |
|
Usually minutes. Automation can minimize delays. | Typically under 1 second |
| Amazon Aurora MySQL-Compatible Edition | Global database has secondary cluster(s) in a different Region |
|
Typically under 1 minute. | Typically under 1 second |
| Amazon Aurora PostgreSQL-Compatible Edition | Global database has secondary cluster(s) in a different Region |
|
Typically under 1 minute. | Typically under 1 second |
| Amazon DocumentDB (with MongoDB compatibility) | Global cluster has secondary cluster(s) in a different Region |
|
Usually minutes. Automation can minimize delays. | Typically in seconds |
| Amazon ElastiCache for Redis | Global datastore has a secondary cluster in a different Region |
|
Usually minutes. Automation can minimize delays. | Typically under 1 second |
| Amazon DynamoDB | DynamoDB global tables | Active/active configuration accepts write operations in secondary Region | Zero or near zero. | Sub-second |
| Amazon RDS for Oracle | Read replica promotion across Regions (Oracle Enterprise Edition and Active Data Guard) |
|
Usually minutes. | Usually minutes |
| Amazon RDS for Oracle (alternative) | Mounted read replica promotion across Regions (Oracle Enterprise Edition) |
|
Usually minutes. | Usually minutes |
| Amazon RDS for SQL Server | AWS Database Migration Service (AWS DMS) requires change data capture (CDC), which incurs a SQL Server license cost for the target instance | Continuous replication | Depends on redo log volume (usually minutes). | Near zero |
