How the Amazon Timehub team built a data replication framework using AWS DMS: Part 1

This article discusses how the Amazon Timehub team built a data replication framework using AWS Database Migration Service (AWS DMS) and Amazon Aurora PostgreSQL-Compatible Edition as the target database. The solution addresses the need for a sub-minute latency, high-performant, single source of truth (SSOT) data store for Amazon's time and attendance data.

Specifically, the article covers:

• The problem statement and business pain points with the existing solution, including high response time for read APIs, lack of a good mechanism for capturing incremental change data, and potential data loss events.
• The approach taken to evaluate AWS DMS as a solution, including benchmarking tests and production workload analysis.
• An overview of the solution architecture, involving AWS DMS, Oracle as the source database, and an Aurora PostgreSQL cluster as the target.
• Key considerations for configuring the source Oracle database and AWS DMS, such as supplemental logging, CDC methods (LogMiner vs. Binary Reader), instance selection, task design, and task merging.
• Monitoring the solution using various AWS CloudWatch metrics.
• The conclusion highlighting the benefits of the solution, including sub-minute data propagation latency, reduced resource demands on the source database, and reduced development time and cost.