Wisea six-minute endpoint at 800 ms
I built the APIs the business runs on, and made the slowest of them answer in 800 ms instead of six minutes.
800 msfrom 6+ minutes
The problem
Several key endpoints sat on Hibernate access patterns that kept timing out in production. The worst of them took more than six minutes to answer a single request.
To scale. At this width, 800 ms is about the thickness of a line.
What I did
I rewrote that data access. I designed new Postgres schemas from scratch for new product features, ran the AWS messaging between services on SQS and SNS, and later led the re-architecture of a core module.
What changed
The six-minute endpoint answers in 800 ms. The APIs carried more than 90,000 requests a day at 99.9% uptime, and the new module made room for the 20% jump in active customers that followed.
The numbers
- 90k
- requests a day
- 99.9%
- uptime
- 20%+
- more active customers, after
Stack
- Node.js
- TypeScript
- Java
- PostgreSQL
- Redis
- AWS SQS
- AWS SNS
- S3
- Docker
- Sentry
- React