Legacy Consolidation to a Distributed AWS Architecture
Pulled scattered legacy systems onto one AWS architecture. Availability settled at 99.8%.
- ROLE
- Full Stack Software Engineer, MXS Solutions
- PERIOD
- 2021 – Present
- SERVICE AVAILABILITY
- 99.8%
- STACK
- AWS, Docker, Kubernetes, Terraform, CI/CD, Vitest, Playwright
This is client work at MXS Solutions. The code and the running systems are private, so there's nothing to link.
THE PROBLEM
Several systems had grown up separately, doing overlapping jobs, with no deployment or testing approach in common. How reliable a request was depended on which of them it happened to hit, and there was nowhere to go to reason about that.
WHAT I BUILT
- Merged the overlapping systems into one distributed architecture on AWS, with the infrastructure in Terraform instead of in someone's memory.
- Put Vitest and Playwright suites in the deploy path. Regressions started showing up before release.
- Moved static and cacheable delivery to a CDN. Less load on origin, shorter path to the user.
- Drew API boundaries between the services so each could deploy and scale on its own.
OUTCOME
- 99.8% availability across the consolidated architecture.
- Deploys became routine. Tests and reproducible infrastructure did the checking a person used to do.
- One system to operate and reason about, where there had been several drifting apart.