Rate Limiting and Throttling

Part of the Web course

At 09:14 your API starts timing out for everybody. Nothing was deployed. It turns out one customer's integration has a retry loop with no delay in it, and a single machine is sending four thousand requests a second to an endpoint that was sized for forty. The system did exactly what you told it to: it tried to serve every request, and it fell over serving them.

Every rate limiter answers one question: has this client had its share? The algorithms differ in how they count. ...