Challenge: Rate-Limited Analytics API with Hot-Key Cache (Spring Boot)

Problem

Build an analytics API in Spring Boot, in-memory only, and keep correctness under parallel load.

Endpoints

1) POST /events

• Body: { "userId": "u1", "type": "click" }
• Validate type in click|view|purchase
• Store event
• Rate limit: 100 requests/minute per userId
• Return 429 if exceeded

2) GET /stats?type=click&window=60

• Return the count of events of type in the last window seconds
• Example response:

{ "type": "click", "window": 60, "count": 123 }

Concurrency & Caching

• Must be concurrency-safe under parallel requests
• Add an in-memory cache for hot stats queries
• Cache key: type + window
• TTL: 5 seconds
• Must not return stale data past TTL
• Avoid thundering herd:
• Only one recompute per key at a time
• Other requests should either wait or serve old value (you decide, document it)

Storage

• In-memory only (no DB)

Acceptance Criteria

• Correct rate limiting under concurrent requests
• GET /stats is accurate for the requested window
• Cache works and is concurrency-safe
• Under parallel load, only rate-limited requests fail

Stretch Goals

• Sliding window counter instead of storing all events
• GET /health
• Basic metrics: total events, cache hits/misses

Expected Deliverables

• Source code
• A short README explaining:
• Chosen data structures
• Cache invalidation and TTL strategy
• Herd-control strategy
• Concurrency/load test evidence