Solution

Sharding (Horizontal Partitioning)

Sharding implies distributing data across multiple physical servers (shards). Each shard holds a subset of the data and acts as a self-contained database.

Sharding Strategies

1. Key-Based (Hash) Sharding: hash(user_id) % num_servers. Good distribution, but hard to add servers (requires re-sharding).
2. Range-Based Sharding: Users A-M go to Server 1, N-Z to Server 2. Easy to query ranges, but leads to uneven distribution.

The Hot Partition Problem

If you shard by "Celebrity User ID" or "Region", one shard might receive 90% of the traffic (e.g., all users querying Justin Bieber's tweets). This defeats the purpose of distributed systems. Consistent Hashing is often used to mitigate this.