Booking flow
Search → availability → hold → confirm, resilient to retries.
Platform · System design · Terraform cloud
A booking flow that stays fast when everyone books at once: cached availability, queued reservations, and cloud infrastructure expressed entirely in Terraform.
Challenge
Flash sales drove traffic spikes that overwhelmed availability checks, causing double-bookings and abandoned carts. Infrastructure was hand-built in the console, so scaling meant late-night clicking.
Approach
We separated availability reads (heavily cached, served at the edge) from reservation writes (queued, idempotent, confirmed asynchronously). The whole cloud footprint moved into Terraform modules with autoscaling policies and load-tested limits.
Search → availability → hold → confirm, resilient to retries.
Edge-cached reads with queued, idempotent reservations.
VPC, autoscaling, budgets, and alerts — all in code.
Game-day tested to multiples of peak-season traffic.
Calm
checkouts through flash-sale spikes
Coded
infrastructure — no more console clicks
Clear
costs with budgets and alerts per service