Amazon Web Services
2483 Case Studies
A Amazon Web Services Case Study
Amazon’s Item Master Service (IMS) ingests and hosts authoritative catalog data for Amazon’s massive selection—processing more than five billion catalog updates a day. Rapid growth (up to 50% annually) and the limits of the Oracle persistence layer (connection/I/O caps, frequent overprovisioning, and failovers that cut availability by up to 50%) forced the engineering team to spend ~40% of its time scaling databases and handling outages for a dataset that had grown to ~600 billion records (~150 TB).
The team migrated IMS to Amazon DynamoDB using AWS Database Migration Service in a two-phase plan (live write switch and backfill), transforming data with DMS templates and using optimistic concurrency to avoid data loss. They moved ~150 TB from 24 Oracle databases in under two months with no downtime. Results included a 90% reduction in operational effort, a 10x improvement in availability, much faster peak-prep time (weeks to hours), simplified architecture, automatic scaling and lower overall cost.