Context
I worked in a high-traffic e-commerce environment after its infrastructure migration from AWS to GCP.
Problem
The system began experiencing database timeouts under higher traffic. It was not obvious which layer was failing: the database, the application, or the new cloud setup.
Investigation
- Used load testing and concurrency analysis to investigate behavior under traffic.
- Examined cold starts and database connection pooling.
- Analyzed caching and autoscaling alongside database and infrastructure bottlenecks.
Root cause
The timeouts were not one misconfigured parameter. They came from the interaction of cold starts, connection pooling, caching and autoscaling after the migration.
My contribution
- Investigated system behavior under load instead of guessing at a fix.
- Tested hypotheses before proposing architecture changes.
- Analyzed database and infrastructure bottlenecks.
- Contributed to improvements involving pooling, caching and autoscaling.
Outcome
Pooling, caching and autoscaling changes followed the investigation, and the system handled traffic more reliably afterwards.
What I learned
Performance problems are rarely solved by changing one parameter. Measure each layer before deciding where to optimize.