05 of 10
Infrastructure
Smart Parking System.
A camera-driven parking system handling 100,000+ vehicles across multiple lots. Zero double-bookings since launch, after a Postgres row-level lock replaced application-layer concurrency.
- Built at
- DevTechGuru
- Scale
- 100,000+ cars, multiple locations.
- Category
- Infrastructure
- Booking Errors
- 0
- Cars Tracked
- 100k+
Context
Security guards were recording licence plates by hand. Queues formed at the gate and revenue was leaking.
Challenge
Sync live camera feeds and payment state in real time, without latency drivers would notice.
Non-negotiables
- No double-bookings
- Video and database always consistent
- Every payment accounted for
Calls I made
- 01
Row-level lock on reservation
A Postgres row-level lock at the moment of reservation prevents the same spot from being sold twice under concurrent load.
- 02
Hardware adapter layer
A pluggable adapter abstracts camera vendors. Hardware swaps do not change application code.
Trade-offs accepted
- Under heavy load the system refuses a booking rather than risk a conflict. Conservative on throughput, zero double-bookings.
- Deferred analytics dashboards until the reservation path was stable.
Outcome
100,000+ vehicle transactions since launch, zero double-bookings recorded.
When software drives a physical action, partial correctness is not enough. The cost of a wrong answer is on the ground.
Stack
- Django
- Next.js
- PostgreSQL
- Row-level locking
- Camera adapter layer
- AWS
Django
Next.js
PostgreSQLGot something similar in mind?
Send 3 lines. I reply within a day.