Case Study: Amazon.com achieves 50 PB Oracle-to-AWS analytics migration and a scalable, cost-efficient data lake with Amazon Web Services (AWS)

Amazon Migrates 50 PB of Analytics Data from Oracle to AWS

Amazon’s consumer business ran a massive analytics operation—publishing more than 50 PB of data across 75,000 tables with 600,000 analytics jobs per day—supported by a monolithic on-premises Oracle data warehouse. The Oracle platform wasn’t built for petabyte-scale processing, suffered frequent failures on large-table transforms, required hundreds of engineer-hours each month for maintenance and upgrades, and imposed high, inflexible licensing and capacity costs that limited analytics and ML work.

Amazon moved to an AWS-based data lake on Amazon S3 and a federated analytics stack (Amazon EMR, Amazon Redshift, Redshift Spectrum and EC2), added metadata and governance services, and used AWS SCT to automate conversion of ~80% of 200,000 queries—saving roughly 1,000 person‑months. The result is a much larger, more flexible environment (the lake grew to ~200 PB) with thousands of Redshift/EMR clusters, eliminated Oracle licensing, retired 30% of unused workloads, lower latency and costs, and broader self-service analytics and engineering focus.

Amazon.com