MockSchool AI

Does your rate limit still hold across three servers?

One skill · 5 minutes · Speak or type · Whiteboard optional

The challenge

Three API servers share traffic for one customer. Enforce at most 100 accepted requests per customer per fixed one-minute window across all three servers. In five minutes, explain where the decision happens, how concurrency is handled, and what happens if the limiter is unavailable. A burst across a fixed-window boundary is permitted in this exercise.

Try a 5-minute design challenge

Try the interview without an account. Create a free account to see your report.

Illustrative example — not your assessment

An invented, synthetic answer to a different challenge: What happens to your feed when delivery falls behind?

Ordinary authors fan out asynchronously. High-follower posts are merged during reads. I track delivery lag and deduplicate the merged result. The target is visibility within a few seconds, not immediate consistency for every reader.

This is an invented example written for this page, not a customer answer.

About this challenge

Read this after you have tried the challenge. It explains one approach.

A limit enforced separately on each server may not enforce the customer’s total limit. Start by defining the key, the window, and what counts as an accepted request. Then explain how simultaneous requests make one shared decision. A shared atomic operation is one option; partitioned quotas can also work if their reservations preserve the total budget. State the outage policy. If your requirement is a strict cap, accepting unlimited requests when the limiter is down breaks it. A fixed-window policy can allow bursts across a boundary, so do not promise a rolling-window limit without a mechanism that enforces one.

Handle two requests at the limit

More challenges

MockSchool AI offers AI-driven interview practice. Feedback is practice feedback on the part of the answer you give, not a prediction of any interview outcome.