Top K System (YouTube Top K)

Difficulty: Hard

What It Tests

Approximate counting at massive scale, heavy hitters algorithm, two-stage aggregation architecture.

Topics Covered

• Distributed Counting / Top K
• Stream Processing
• Consistent Hashing
• Caching
• Redis Deep Dive

Hello Interview Breakdown

Read the full Hello Interview breakdown →

Video Walkthroughs

• ByteByteGo: Top-K Problem (Heavy Hitters) System Design →
• System Design Interview: Design Top-K YouTube Videos →

Approach Hints

• Two-stage architecture: local Count-Min Sketch per ingest shard counts events; merge local top-K lists into a global top-K every few seconds
• Store the global top-K in a Redis sorted set (ZREVRANGE) for O(log N) updates and O(k) reads
• Use a sliding window (last 1h, 24h, all-time) — each window requires a separate aggregation pipeline
• Approximate algorithms (Space-Saving) allow O(k) memory regardless of the total number of distinct items