User Likes: Design a Scalable Likes and Reactions System for a Social Platform

Question

## Round 1 - System Design

Design the backend for a "likes" system on a social platform at Facebook/Instagram scale.

**Requirements:**
- Users can like or unlike posts, comments, and stories.
- Show the like count on each item (eventually consistent is acceptable).
- Show whether the current user has liked a given item (must be accurate).
- Top-liked posts in the last 24 hours (leaderboard).
- Scale: 500M DAU, 100K like events/second.

## Data Model

```sql
CREATE TABLE likes (
  user_id   BIGINT,
  item_id   BIGINT,
  item_type TINYINT,  -- 0=post, 1=comment, 2=story
  liked_at  TIMESTAMP,
  PRIMARY KEY (user_id, item_id, item_type)
);
```

**Like count:** Store in a Redis counter per item. Periodically flush to DB.
**Has-liked check:** Query the likes table directly (indexed on user_id + item_id).

## Follow-ups

1. How do you handle a viral post receiving 1M likes/second? Does your Redis key become a hot spot?
2. How do you ensure the like count in Redis and in the DB don't diverge permanently on a crash?
3. What schema changes are needed to support reaction types (love, haha, wow) instead of just likes?
4. How would you build the 24-hour top-liked leaderboard efficiently?

User Likes: Design a Scalable Likes and Reactions System for a Social Platform

Question Details

Full Details

Topics

More from Roblox