Amazon’s consumer business says it has turned off its final Oracle database, after migrating a massive 75 petabytes of internal data stored on nearly 7,500 Oracle databases to a range of its own AWS services.

The mammoth migration involved over 100 internal Amazon teams, from across Alexa, Amazon Prime Video, Kindle, Amazon Music, and Twitch, as well as internal teams spanning advertising tech, payments, infosec and retail.

The move comes 14 months after Oracle’s Larry Ellison took a jab at Amazon’s reliance on his company’s infrastructure, saying “it’s kind of embarrassing when Amazon uses Oracle but they want you to use Aurora and Redshift.”

“They’ve had 10 years to get off Oracle, and they’re still on Oracle.”

Amazon said in a blog today that it had reduced its database costs by over 60 percent “on top of the heavily discounted rate we negotiated based on our scale.”

It also said it had improved the speed of its consumer-facing applications by 40 percent, and reduced database administration overheads by 70 percent as a result of switching to its own managed services, including Amazon DynamoDBAmazon AuroraAmazon Relational Database Service (RDS), and Amazon Redshift.

“The migrations were accomplished with little or no downtime, and covered 100% of our proprietary systems. This includes complex purchasing, catalog management, order fulfillment, accounting, and video streaming workloads,” AWS’ Jeff Barr wrote.

See also: Sports App Strava Ditches AWS Redshift for Snowflake

The company is planning a series of sessions at its annual Re:Invent summit in Las Vegas on the migration effort, it said, saying it had trained its Oracle DBAs up as database migration specialists and advisors, with their old roles now redundant.

Barr said Amazon’s advertising team in particular was able to double its database fleet size (and their throughput) in minutes to accommodate peak traffic, courtesy of RDS in a scale-up effort that would have previously “taken months”, while its financial ledger team moved 120 TB of data, reduced latency by 40 percent, cut costs by 70 percent, and cut overhead by the same 70 percent by using DynamoDB.

This costs and claims are, of course, hard to cross-check: readers will have to take AWS’s word for it and the company has more than a slight vested interest in telling a story of entirely smooth sailing. Oracle, meanwhile, is pushing heavily into the cloud and says it will have 36 Oracle Cloud Infrastructure regions by the end of 2020; more than AWS.

See also: The Life of a DBA: A Peek Behind the Curtain into Stress, Cost-Cutting and (Sometimes) Chaos