Design multi-region active-active applications with DynamoDB Global Tables. Understand the difference from Aurora Global Database.
DynamoDB Global Tables — Multi-Region Active-Active
DynamoDB Global Tables provide fully-managed multi-region, multi-active database replication. Unlike Aurora Global Database (one primary, multiple read-only secondaries), DynamoDB Global Tables are ACTIVE-ACTIVE — you can write to ANY region.
Teacher Note: Aurora Global = one chef (primary region), multiple servers (read replicas). DynamoDB Global Tables = multiple chefs in multiple countries, all taking orders simultaneously, and constantly sharing their order book with each other.
How Global Tables Work
Global Table: 'users-table'
US Region (us-east-1):
Table replica
Accepts READS and WRITES
Replicates changes to: EU, APAC
EU Region (eu-west-1):
Table replica
Accepts READS and WRITES
Replicates changes to: US, APAC
APAC Region (ap-southeast-1):
Table replica
Accepts READS and WRITES
Replicates changes to: US, EU
Replication: sub-second across all regions
Conflict resolution: 'last writer wins' (timestamp-based)
Use Cases
- Gaming leaderboards: players worldwide write scores locally, sub-second global consistency
- User profiles: write profile updates to nearest region, read locally everywhere
- IoT device state: devices write to nearest region, global view of device state
- Collaboration tools: users in different regions write simultaneously
Global Tables vs Aurora Global Database
| Feature | DynamoDB Global Tables | Aurora Global Database |
|---|
| Write regions | ALL regions (active-active) | ONE primary region only |
| Read regions | All regions | All regions |
| Conflict resolution | Last writer wins | Not applicable (one writer) |
| Use case | True multi-region writes | Global reads + DR failover |
| Data model | NoSQL (key-value/document) | SQL (relational) |
Exam Tip: DynamoDB Global Tables are the best choice when you need true multi-region active-active writes. The 'last writer wins' conflict resolution means simultaneous writes to the same item in different regions — the one with the latest timestamp wins. Design your data model to minimise cross-region write conflicts.