Personal Project
Mini Scalable Admission System
A Laravel 12 + Next.js proof of concept for the problem real admission systems actually have: not total data volume, but traffic bursts when results drop and everyone arrives in the same few minutes. It implements queued writes, row-level seat locking and read-through caching, then measures with k6 whether they help — keeping the naive version on a branch so the comparison has something honest to run against.
- Queued writes return 202 Accepted before the row exists — p99 latency down 65% below saturation
- SELECT ... FOR UPDATE seat allocation that cannot oversubscribe: 50 racing processes, 20 seats, 20 handed out
- Read-through Redis cache for result lookups, hitting Postgres only on a miss (X-Cache: HIT/MISS)
- k6 baseline-vs-queued comparison — and the published measurement where the queue loses
- 79 tests against real Postgres, not SQLite, because three of them passed for the wrong reason on SQLite
Laravel 12Next.jsPostgreSQLRedisDockerk6
The API is on Render's free tier and sleeps after 15 minutes idle — the first request wakes it and can take ~25 seconds.
