Patrick Collison

Stripe achieved 99.99986% API uptime with MongoDB and Ruby as core tech

Patrick Collison reveals how Stripe's early technology choices of MongoDB and Ruby became foundational to the company, despite requiring significant infrastructure investment. He explains how they built custom systems to make MongoDB fault-tolerant and distributed enough for their needs, ultimately achieving exceptional API reliability of 99.99986%.

Stripe had to rewrite key services from Ruby to Java for performance

Patrick Collison discusses Stripe's technical evolution, revealing how the company had extensive internal documentation about migrating from Ruby to Java. He explains that they partially executed this migration, rewriting critical services in Java when throughput performance became essential, acknowledging Ruby's limitations under heavy load.

Stripe's 12-year API evolution from v1 to their major 2022 redesign

Patrick Collison explains how Stripe has been planning for API evolution since 2010, prefixing all REST URIs with '/v1' from the beginning. He discusses their decision in 2022 to finally increment their API namespace and ship new functionality that enables capabilities beyond their original architecture.

Stripe realized their core APIs were fundamentally wrong and had to rebuild

Patrick Collison reveals how Stripe discovered in 2022 that some of their core data abstractions were fundamentally flawed for long-term growth, forcing them to design v2 APIs. He explains how they fortunately planned for this possibility from the beginning by versioning their REST APIs with '/v1' prefixes since 2010, making the transition manageable.

Stripe achieved 99.99986% API uptime - just 44 seconds down all year

Patrick Collison reveals Stripe's exceptional infrastructure reliability, sharing that their API was unavailable for only 44 seconds in an entire year. He discusses the significant engineering investment required to make MongoDB fault-tolerant and distributed enough to achieve what he believes is industry-leading uptime performance.

in our technology choices than our prior company. And so instead of using small talk, you know, okay. We aren't gonna go to Java, but we went to Ruby, which at least on a relative basis seemed more...