Database Sharding Strategies

Sharding splits a large database into smaller, faster, distributed pieces called shards.

Why Shard?

• Single server has storage limits
• High write throughput requirements
• Reduce query latency

Sharding Strategies

Range-Based Sharding

Shard 1: user_id 1 – 1,000,000
Shard 2: user_id 1,000,001 – 2,000,000
Shard 3: user_id 2,000,001 – 3,000,000

Pros: Simple, good for range queries Cons: Hotspot problem (new users always hit last shard)

Hash-Based Sharding

Shard = hash(user_id) % num_shards

Pros: Even distribution Cons: Range queries span all shards

Directory-Based Sharding

Lookup table maps keys to shard.

Pros: Flexible shard assignment Cons: Lookup table is single point of failure

Resharding

When a shard is too large:

1. Split shard into two
2. Update routing logic
3. Migrate data

Challenges

• Cross-shard joins are expensive
• Distributed transactions are complex
• Rebalancing requires data migration

Conclusion

Hash sharding is the most common for even distribution. Combine with replication for high availability.