Amazon DynamoDB Global Tables Specialized2017年〜
A feature that replicates DynamoDB tables across multiple regions in an active-active configuration
What It Does
Amazon DynamoDB Global Tables automatically replicates DynamoDB tables across multiple AWS regions. It enables an active-active configuration where reads and writes are possible in each region, providing low-latency data access for global applications.
Use Cases
It is used for data synchronization in globally deployed mobile apps and games, multi-region disaster recovery, and providing low-latency access to geographically distributed users.
Everyday Analogy
Think of it like a bank with branches worldwide. You can check and update your account balance (data) at any branch (region), and changes are automatically synced across all branches.
What Are Global Tables?
DynamoDB Global Tables provides multi-region replication for tables. Simply specify the regions where you want to add replica tables, and data is automatically synchronized bidirectionally. Each regional replica supports both reads and writes, with data typically propagating within 1 second.
Conflict Resolution and Consistency
When the same item is updated simultaneously in multiple regions, Global Tables resolves conflicts using a 'Last Writer Wins' approach. The update with the most recent timestamp becomes the final value. Designing your application to avoid conflicts (such as region-specific data partitioning) is also effective. For a systematic understanding of conflict resolution and consistency, books (Amazon) provide comprehensive coverage.
Getting Started
Select an existing table in the DynamoDB console, go to the 'Global tables' tab, and click 'Create replica' to add a region. DynamoDB Streams is automatically enabled and replication begins. Adding and removing replicas can be done online with no downtime.
Things to Watch Out For
- Write capacity unit charges apply per replica region. Data transfer charges for replication are also added
- Conflict resolution uses Last Writer Wins, so design considerations are needed for workloads with frequent concurrent updates to the same item
