Migrating from Oracle to Amazon Aurora in 18 weeks with The New York Times
Learn how The New York Times modernized its digital subscription system by migrating from Oracle to Amazon Aurora PostgreSQL.
Benefits
- lower projected database total cost of ownership
-
26%
- month migration compressed into 18 weeks
- 18
- reduction in database-related operational work
- 36%
Overview
Since introducing its digital paywall, The New York Times has grown into a subscription-first news organization, delivering journalism to readers across every device and platform. But PLATO, the in-house system that processes these subscriptions and authorizes reader access, was tied to a legacy Oracle database that couldn’t affordably scale to meet modern demands. Using Amazon Web Services (AWS), the company modernized that mission-critical system, migrating it from Oracle to Amazon Aurora PostgreSQL-Compatible Edition—a relational database service that provides unparalleled high performance and availability—in 18 weeks instead of the projected 18 months. The migration eliminated maintenance downtime and is projected to reduce the total cost of ownership of the database by 26 percent.
About The New York Times
The New York Times is a US-based news organization that serves a global audience through its website, mobile apps, and syndication partners.
Opportunity | Scaling access using AWS for The New York Times
When The New York Times launched its paywall in 2011, no commercial product could meet the needs of its business, so its engineers built PLATO, a digital subscription system, in house. The New York Times’ entire network draws its entitlement decisions from this backend, which ran on a self-managed Oracle database hosted on Amazon Elastic Compute Cloud (Amazon EC2), a service that provides resizable, secure compute capacity.
Over time, this arrangement grew harder to sustain. Oracle licensing is priced per processor core, so scaling out with more compute or reader instances would have multiplied licensing costs. Billing jobs and subscriber lookups competed for the same resources, and a spike in one slowed the other. Twice a year, maintenance took the system offline for at least 15 minutes. To overcome these limitations, The New York Times decided to use AWS.
Solution | Modernizing a mission-critical subscription system
The New York Times engaged Mactores Cognition (Mactores), an AWS Partner, and participated in the AWS Migration Acceleration Program (AWS MAP), which accelerates cloud migration and modernization journeys, to migrate from the Oracle database to Amazon Aurora PostgreSQL-Compatible Edition. The teams determined that a migration using conventional methods would take 18 months. The 15-year-old system included approximately 110,000 conversion items across three schemas and more than 8,000 application files with nearly 6 million lines of code, and Oracle SQL was embedded throughout Java, Python, Go, Groovy, and shell scripts.
However, the Mactores Aedeon Agent—an AI migration agent built on Amazon Bedrock, the solution for building generative AI applications and agents at production scale, with Anthropic’s Claude Opus 4.6—could compress that work by analyzing the codebase and automating code conversion. “With the new agentic AI tools, the AWS team, and our vendor partner, we moved an 18-month timeline into 18 weeks,” says Joe Cora, senior engineering manager at The New York Times.
The New York Times chose Aurora PostgreSQL as the foundation for PLATO. Using Aurora, the company could separate its reads from its writes across up to 15 read replicas, improving read scalability while keeping the primary instance free for writes. The New York Times already used PostgreSQL as the internal standard for relational workloads, so its engineers could work with familiar database technology because Aurora is fully compatible with PostgreSQL.
Using AWS migration services and deep AWS expertise, the teams migrated the database and modernized the underlying application code in the same pass. To structure the work, The New York Times used AWS Database Migration Service (AWS DMS) to keep the source system fully operational until the migration was complete. AWS DMS also maintained continuous replication to a clone of the production Oracle database, which gave the team a rollback path with zero risk of data loss. The team also used AWS DMS Schema Conversion to automate the vast majority of database object conversion, while the Aedeon Agent converted the SQL embedded in application code. Before cutover, the team mirrored live PLATO traffic to run shadow queries against PostgreSQL behind feature flags. This surfaced query-plan and edge-case issues while Oracle remained the system of record.
Outcome | Achieving zero maintenance downtime and lower costs
PLATO went live on Aurora PostgreSQL 18 weeks after work began. AI conversion and automatic generation through an object-relational mapping framework handled 84 percent of the SQL, so engineers hand-edited only 16 percent. At launch, the system sustained 18,500 requests per second with 100 percent row-level data accuracy and zero revenue impact at cutover. The new architecture runs one writer and two reader instances, so the team can route high-volume read traffic, such as subscription and entitlement lookups, away from transactional writes, such as billing and new subscriptions.
After going live, The New York Times adopted Amazon RDS Proxy, a fully managed, highly available, and simple-to-use database proxy, which improves application scalability and resilience by pooling and sharing database connections. The team also activated AWS DevOps Agent, a frontier agent for release management and production operations, to help identify and mitigate database performance issues.
“Since migrating to Amazon Aurora PostgreSQL, we’ve gone through a full maintenance window with zero downtime,” says Cora. “We can keep serving subscribers and taking new subscriptions, even during maintenance.”
The New York Times has reduced database-related operational work by 36 percent and expects to eliminate an estimated 15 Oracle-specific operational tickets each year. By removing that recurring maintenance work and Oracle licensing costs, the company projects a 26 percent lower total cost of ownership for the database over 5 years. Next, The New York Times plans to scope an AWS multi–Region deployment to extend the same resilience and availability across the geographies that it serves.
Since migrating to Amazon Aurora PostgreSQL, we’ve gone through a full maintenance window with zero downtime. We can keep serving subscribers and taking new subscriptions, even during maintenance.
Joe Cora
Senior Engineering Manager, The New York TimesAWS Services Used
More Customer Stories
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages