Implement a Per-IP Sliding Window Rate Limiter
Company: Ramp
Role: Software Engineer
Category: Software Engineering Fundamentals
Difficulty: medium
Interview Round: Technical Screen
Given sorted request timestamps, matching IP addresses, a request limit, and a time-window length, return an array where `1` means the request is accepted and `0` means it is rejected. For each IP, at most `limit` accepted requests may occur within any window of length `timeWindow`.
### Constraints & Assumptions
- Timestamps are sorted in nondecreasing order.
- The window is per IP address.
- Rejected requests should not count against future accepted-request limits.
- Return one decision per input request.
### Clarifying Questions to Ask
- Is the time window inclusive on both ends or half-open?
- Should rejected requests be tracked at all?
- Can multiple requests from the same IP share a timestamp?
- What are the input size limits?
- Are IPs strings or numeric ids?
### What a Strong Answer Covers
```premium-lock What a Strong Answer Covers
```
### Follow-up Questions
- How would this run in a distributed system?
- How would you implement token bucket instead?
- How would memory be bounded for many inactive IPs?
- How would you explain rejected versus accepted request accounting?
Quick Answer: This Ramp software engineering question evaluates implementation of a per-IP sliding window rate limiter. It prepares candidates to discuss time-window data structures, memory cleanup, concurrency concerns, and operational behavior under traffic spikes.