🛒 Flash sale system design
A simplified, healthy design: 13 app servers behind a load balancer, a cache (90% hit rate), an async write queue, a CDN for static files, a primary database. At 5,500 requests per second it runs at p95 46 ms with 0.0% errors for about $730/month.
Healthy at 5,500 req/s
- p95 latency (server)
- 46 ms
- Error rate
- 0.0%
- Est. cost / month
- $730
- excl. bandwidth
App servers
400 req/s each · $40/mo
Hit rate
DB read replicas
+1,250 reads/s each · also applies every write · $120/mo
Primary DB: 1,250 queries/s (1 query per request). How the numbers work
How it works
- Everyone arrives at the same second, and many of them try to buy (write) at once.
- A queue turns the stampede into an orderly line: users get a fast "you are in line" while orders are written in the background, as long as the database keeps up on average.
- Product pages and images are cached and served from the CDN so browsing stays fast.
Drop the cache hit rate to 50%. Every miss lands on the one primary database.
Use the Build and Traffic tabs above. The overloaded part turns red and starts dropping requests.