PayPal’s Migration from Java to Node.js: Performance Gains and Development Efficiency

PayPal announced that it is moving its large‑scale web applications from Java to Node.js. The migration was motivated by a claimed two‑fold increase in development efficiency, a two‑fold performance boost, a 33% reduction in code volume, and a 40% decrease in the number of files.

The shift is also driven by concerns that Java is becoming increasingly closed and complex, with Oracle planning to charge for JVMs, prompting many companies to “de‑Java” their stacks.

Historically, PayPal’s engineers were split between front‑end (HTML, CSS, JavaScript) and back‑end (Java) teams. By adopting a full‑stack approach, the same engineers can now handle both UI and server logic, eliminating the bottleneck of coordinating separate teams.

Initial experiments used Express for routing and nconf for configuration, but the team found Express insufficiently scalable for large‑team collaboration. They therefore built an internal framework called Kraken.js, which functions more as a convention than a traditional framework and better supports large‑scale development.

The first production Node.js service was the company’s most‑visited home page. To mitigate risk, the Node.js version ran in parallel with the existing Java implementation, allowing an immediate fallback if issues arose while collecting performance data.

After two months of development, both the Java and Node.js teams delivered identical functionality. Unit‑test results showed that the Node.js application required fewer developers, reduced code size by 33%, and cut the number of files by 40% compared with the Java counterpart.

Based on these results, PayPal decided to pause Java development and focus entirely on JavaScript‑based applications, a move welcomed by engineers who had previously been skeptical of Node.js.

Performance testing compared a Spring‑based Java service with a stack built on Kraken.js, Express, and Dust.js. The Node.js service handled roughly twice the requests per second and rendered pages 35% faster (about 200 ms saved per page), even though the test used a single‑core Node.js process against a five‑core Java process.

Looking ahead, PayPal plans to continue building its web applications with Node.js, including new portals, user‑overview pages, and several beta projects, and will share further lessons and data from the migration.