Question #1254
A company is migrating its on-premises Oracle database to Amazon Aurora PostgreSQL. The database supports multiple applications that concurrently access the same tables. Applications must be migrated weekly over a six-week period, and the database experiences heavy read and write operations. Management requires continuous data synchronization between both databases during the migration to ensure consistency. What should a solutions architect recommend?
Use AWS DataSync for the initial data transfer. Configure AWS Database Migration Service (AWS DMS) with a change data capture (CDC) replication task and table mappings for all tables.
Use AWS DataSync for the initial data transfer. Configure AWS Database Migration Service (AWS DMS) with a full load plus change data capture (CDC) replication task and table mappings for all tables.
Use the AWS Schema Conversion Tool with AWS Database Migration Service (AWS DMS) on a memory-optimized replication instance. Configure a full load plus change data capture (CDC) replication task with table mappings for all tables.
Use the AWS Schema Conversion Tool with AWS Database Migration Service (AWS DMS) on a compute-optimized replication instance. Configure a full load plus change data capture (CDC) replication task with table mappings for the most frequently accessed tables.
Explanation
Option C is correct because:
1. AWS Schema Conversion Tool (SCT) is required to convert Oracle schemas to PostgreSQL-compatible schemas, ensuring structural compatibility.
2. AWS DMS with full load + CDC performs the initial data migration (full load) and continuously replicates ongoing changes (CDC), meeting the requirement for continuous synchronization.
3. Memory-optimized replication instance is ideal for heavy read/write workloads, ensuring minimal latency during replication.
Why other options are incorrect:
- A/B: DataSync is designed for bulk file transfers, not live database migration. DMS alone handles initial load and CDC.
- D: Using table mappings only for frequently accessed tables risks data inconsistency. Compute-optimized instances are better for CPU-heavy tasks, while memory-optimized suits high-throughput database operations.
Key Points: Use SCT for schema conversion, DMS full load + CDC for continuous sync, and memory-optimized instances for heavy workloads.
Answer
The correct answer is: C